Wei Zhuang, Tomoyuki Hayashi, and Shaul Mukamel*

Spectroscopic MethodsDOI: 10.1002/anie.200802644

Coherent Multidimensional Vibrational Spectroscopy ofBiomolecules: Concepts, Simulations, and ChallengesWei Zhuang, Tomoyuki Hayashi, and Shaul Mukamel*

AngewandteChemie

Keywords:chirality · molecular dynamics · proteins ·spectroscopic methods ·vibrational spectroscopy

S. Mukamel et al.Reviews

3750 www.angewandte.org � 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Angew. Chem. Int. Ed. 2009, 48, 3750 – 3781

http://www.angewandte.org

1. Introduction

The structure and function of biomolecules are intimatelyconnected; this is one of the central principles of structuralbiology.[1] Predicting protein structures requires understand-ing the interactions and driving forces which cause them tofold from a disordered, random-coiled, state into a uniquenative structure. Exploring the folding mechanism in detailrequires techniques that can monitor the structures withadequate temporal and spatial resolution. X-ray crystallog-raphy can determine the static structure with atomic reso-lution.[2] Time-resolved small-angle X-ray scattering givesmainly the radius of gyration with up to picosecond timeresolution.[3–6] Three-dimensional atomic-resolution struc-tures can be determined using NMR spectroscopy[7,8] butonly on time scales longer than the radiowave period(microsecond).[8] Higher temporal resolution is required formonitoring many elementary biophysical processes, forexample, the a-helix formation,[9] which occurs on a timescaleof hundreds of nanosecond. Nanosecond to picosecondprocesses may sometimes be probed through the frequencydependence of NMR relaxation rates.[10] Such measurementsare indirect and their interpretation is model-dependent.Time-resolved X-ray diffraction provides picosecond snap-shots of structures in crystals.[11] Ultrafast electron pulses arebeing developed as well for time-resolved electron diffractionapplications.[12, 13]

Over the past decade, time-resolved infrared spectroscopycarried out with 20–100 fs laser pulses has emerged as apowerful tool in the investigation of protein folding,[14] thanksto fast laser triggering and the fairly localized nature ofvibrational transitions.[14–20] The coherent techniques surveyed

herein record the molecular response to sequences of pulses,and provide a multidimensional view of their structure.Multidimensional optical techniques are analogues of theirNMR counterparts but with greatly improved temporalresolution.[21, 22] This and other differences are summarizedin Table 1.[23–25] NMR experiments are performed with intensepulses which manipulate the entire spin population. Pulsesequences involving hundreds of pulses are then possible.2DIR studies use weak pulses which only excite a smallfraction of the molecules. Unlike NMR spectroscopy, multipleintense IR pulses can trigger various photophysical andphotochemical processes which are interesting on their own,but complicate the spectroscopic analysis. Therefore, inpractice, only a few weak pulses are used, and the signalscan be calculated perturbatively order by order in theincoming fields. The directionality of the signal (phasematching) stems from the fact that the sample is much

The response of complex molecules to sequences of femtosecondinfrared pulses provides a unique window into their structure,dynamics, and fluctuating environments. Herein we survey thebasic principles of modern two-dimensional infrared (2DIR)spectroscopy, which analogous to those of multidimensional NMRspectroscopy. The perturbative approach for computing thenonlinear optical response of coupled localized chromophores isintroduced and applied to the amide backbone transitions ofproteins, liquid water, membrane lipids, and amyloid fibrils. Thesignals are analyzed using classical molecular dynamics simu-lations combined with an effective fluctuating Hamiltonian forcoupled localized anharmonic vibrations whose dependence on thelocal electrostatic environment is parameterized by an ab initiomap. Several simulation methods, (cumulant expansion of Gaus-sian fluctuation, quasiparticle scattering, the stochastic Liouvilleequations, direct numerical propagation) are surveyed. Chirality-induced techniques which dramatically enhance the resolution aredemonstrated. Signatures of conformational and hydrogen-bonding fluctuations, protein folding, and chemical-exchangeprocesses are discussed.

From the Contents

1. Introduction 3751

2. History of MultidimensionalVibrationalSpectroscopy 3754

3. Hamiltonian Operators for theAmide Vibrations of Polypeptides 3756

4. Liouville-Space Pathways forCoupled Localized Vibrations 3760

5. Spectral Diffusion and ChemicalExchange:The Stochastic Liouville Equations 3764

6. The OH Stretch Band of LiquidWater 3766

7. Application to Phospholipids:Quasiparticle Representation of2DIR Signals 3769

8. Double-Quantum-CoherenceSpectroscopy 3772

9. Chirality Effects: Enhancing theResolution 3772

10. The Structure of Amyloid Fibrils 3774

11. Summary and Outlook 3776

[*] Dr. W. Zhuang, Dr. T. Hayashi, Prof. S. MukamelDepartment of Chemistry, University of California at IrvineCA 92697-2025 (USA)Fax: (+ 1)949-824-4759E-mail: [email protected]

Multidimensional Vibrational SpectroscopyAngewandte

Chemie

3751Angew. Chem. Int. Ed. 2009, 48, 3750 – 3781 � 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

larger than the optical wavelength. In NMR spectroscopy theopposite limit holds: The signal is isotropic. However, thedirectional information may be retrieved by modifying thephases of the pulses (phase cycling). The anharmoniceffective Hamiltonian necessary for 2DIR simulations iscomplex and requires extensive electronic-structure calcula-tions. In contrast, the spin Hamiltonians is NMR spectroscopyare known and universal, greatly simplifying the simulationsand analysis of signals. The dipole moments in NMRspectroscopy are aligned in parallel by the strong magneticfield. In contrast, the specific orientations of IR dipoles carryuseful structural information that can be retrieved by varyingthe pulse polarizations. NMR spectroscopy has a remarkablestructural resolution unmatched by IR signals. However,2DIR provides a different window with complementaryinformation.

A heterodyne-detected 2DIR experiment (Figure 1)involves the interaction of three laser pulses with wavevectors k1, k2, k3, (in chronological order) with the peptide.The coherent signal field is generated along one of the phase-

matching directions: ks =� k1�k2� k3 where all moleculesemit in phase and are detected by interference with a 4th“local-oscillator” pulse with the desired wavevector ks. Thesignal S(t3,t2,t1) is given as the intensity change of the local-

Wei Zhuang studied chemistry at the Univer-sity of California at Irvine until 2007, he iscurrently a postdoctoral fellow at Universityof California at Berkeley.

Tomoyuki Hayashi was born in Tokyo in1973 and earned his PhD with Prof. Hiro-oHamaguchi in 2002. He was an assistantspecialist at University of California, Irvineuntil 2008. He is currently working onelectron transfer in proteins at University ofCalifornia, Davis with Professor AlexeiStuchebrukhov.

Table 1: Comparison of coherent NMR and IR techniques.

NMR IR

Frequency MHz 1013–1014 HzTime Resolu-tion

Millisecond Femtosecond

Hamiltonian Spin Hamiltonian. Few universal parameters. Easier to invertspectra to get structures

Anharmonic vibrational Hamiltonian. Requires electronic structurecalculation. Many parameters. Inversion of signals is more complex

TransitionDipoles

All dipoles of the same nucleus are equal and aligned. Pulsepolarizations and spin states transform by rotating the sample

Varying dipole strength and orientation. Many independent param-eters for the dipole

Pulse inten-sity

Strong saturating pulses. All spins excited, multiple pulsesequences possible

Weak, only few molecules are excited, sequences with few pulsespossible

Modeling The Bloch picture Susceptibilities and response functions

Directionalityof signal

Wavelength l @ sample size, kr ! 1, Signal is isotropic inspace. Pathway selection by phase cycling

l ! sample size, kr @ 1, signal is highly directional. Pathwayselection by spatial phase matching

Targetdegrees offreedom

Spins Molecular vibrations

Temperature High compared to frequencies. Simplifies calculations Low compared to frequencies. Calculation more complicated

Phase con-trol of pulses

Easy Becomes harder as the frequency is increased

Figure 1. Pulse configuration for a heterodyne detected multidimen-sional four-wave mixing experiment. Signals are recorded versus thethree time delays t1, t2, t3, and displayed as 2D correlation plotsinvolving two of the time delays, holding the third time delay fixed [seeEq. (14)].




oscillator field induced by the interactions with the otherfields. Its parametric dependence on the time intervalsbetween pulses carries a wealth of information. 2DIR signalsare typically displayed as two-dimensional correlation plotswith respect to two of these intervals, say t1 and t3, holding thethird (t2) fixed. Since such plots are highly oscillatory, thesignal is double Fourier transformed with respect to the twodesired time variables to generate a frequency–frequencycorrelation plot such as S(W1,t2,W3) where W1 and W3 are thefrequency conjugates to t1 and t3 (Figure 1).

Heterodyne detection allows the whole signal field (bothamplitude and phase) to be recorded. Thus both the real (in-phase) and the imaginary (out-of-phase) components of theresponse can be displayed. Coupled vibrational modes shownew resonances (cross-peaks) whose intensities and profilesgive direct signatures of the correlations between transitions.These are background-free features that vanish for uncoupledvibrations. Correlation plots of dynamic events taking placeduring controlled evolution periods can be interpreted interms of multipoint correlation functions. These carry con-siderably more information than the two-point functions oflinear spectroscopy, and therefore have the capacity todistinguish between possible models whose 1D responsesare virtually identical.

Figure 2, shows simulated 2D photon-echo spectra of twocoupled vibrations. The diagonal peaks at (�2000,2000) and(�2100,2100) resemble the linear absorption. The cross-peaksat (�2000,2100) and (�2100,2000) reveal the couplingsbetween the two modes. The 2D line shapes are very sensitiveto time scales and the degree of correlation of frequencyfluctuations and provide valuable information about thefluctuating environment. Fast fluctuations (Figure 2, rightpanel) show circular diagonal peaks (homogeneous broad-ening) whereas slow fluctuations (Figure 2, left panel) yieldelongated line shapes. In addition the left and middle panelsof Figure 2 show strong variations of the cross-peaks with thedegree of correlation. Bandshape analysis of 2D photonechoes of solute–solvent complexes showed a longer timescale for the slowest of two components in mixed solventsthan in a pure solvent. This result was ascribed to compositionvariations of the first solvent shell.[26] Such fine details are notavailable from 1D measurements.

Pump–probe (also known as transient absorption) is thesimplest nonlinear experiment, both conceptually and tech-nically since it only involves two laser pulses: the pump andthe probe, and requires no phase-control. Typically, the two

pulses are temporally well separated. The system firstinteracts with the pump pulse then with the probe pulse.The difference between the probe transmission with andwithout the pump, reveals information about structuralchanges and energy transport taking place during the delaybetween the two pulses. The Photon-echo signal generated inthe direction ks =�k1 + k2 + k3 is another widely used tech-nique.[28] The excitations generated in the molecule during t1

and t3 acquire an opposite phase, exactly canceling inhomo-geneous broadening in the signal and opening a window intomotions and relaxation timescales. This is not possible with1D techniques, such as linear absorption. Frequency-domainexperiments involving longer pulses, which combine IR andRaman techniques have been carried out.[29]

2D techniques have been widely applied towards explor-ing the equilibrated structures of biomolecules, monitoringpeptide folding dynamics, studying the hydrogen-bondingstructure and dynamics in liquid water, monitoring theelectrostatic environment and its fluctuations around achromophore, or investigating vibrational energy transferpathways. A brief survey of these applications is presented inSection 2.

Fast peptide-folding (in the pico- to nanosecond range)has been extensively studied by Monte Carlo (MC) andmolecular dynamics (MD) simulations.[30–37] Simple latticemodels[35, 36] help develop the big physical picture of thefolding events, while all-atom molecular dynamics simula-tions[32, 37] provide more realistic and detailed informationabout structures and dynamics. Computational powerrestricts such calculations to trajectories of a few tens ofnanoseconds. A 1998 study reported the protein folding withexplicit representation of water for 1 microsecond.[38] Thedirect simulation of protein folding is usually too demand-ing.[39–41] However, MD simulations are gradually acquiringthe capacity to unravel the folding mechanism of peptides and

Shaul Mukamel, the Chancellor Professor ofChemistry at the University of California,Irvine, is the author of over 600 publicationsand the textbook Principles of NonlinearOptical Spectroscopy. He works on designingnovel laser pulse sequences to study molec-ular structure, fluctuations, and energy andchargetransfer processes.

Figure 2. Top: 2D photon-echo spectra of two coupled vibrations inthe phase-matching direction kI =�k1 + k2 + k3. W1 and W3 are theFourier conjugate variables to t1 and t3. The frequency fluctuations ofthe two modes are slow and anti-correlated (left panel), slow andcorrelated (middle panel), fast and anti-correlated (right panel).Adapted from ref. [27]. Bottom: Linear absorptions for the threemodels.


Chemie

3753Angew. Chem. Int. Ed. 2009, 48, 3750 – 3781 � 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim www.angewandte.org


small proteins, thanks to the design of simple model peptidesthat mimic protein complexity, yet are sufficiently small toallow detailed simulations[33, 40, 42–44] and the development andimplementation of powerful simulation algorithms[45] withimproved sampling of rare events.[46,47] Since the visualizationof folding processes strongly depends on simulations, it ishighly desirable to perform experiments on time scalesaccessible to computer simulations. 2DIR and atomisticlevel MD simulations can be carried out on the sametimescale. Thus developing MD methods for simulating2DIR signals can help assign 2DIR features, and unravelthe underlying motions. At the same time the quality ofdifferent MD force fields can be tested by comparing thepredicted 2DIR signatures of different folding pathways withexperiment.

The state-of-the-art computational techniques currentlyemployed in the modeling of 2DIR signals of biomoleculeswill be surveyed in this Review.[48–55] The amide vibrations ofpeptides,[55–57] can be described by the Frenkel exciton modeloriginally developed to describe coupled localized transitionsof oligomers or polymers made out of similar repeat units.The necessary parameters can be obtained from electronic-structure calculations of individual chromophores, rather thanthe whole system, greatly reducing the computationaldemands. The spectrum consists of well-separated bands ofenergy levels representing single excitations, double excita-tions, and higher excitations. The molecular Hamiltonianconserves the number of excitations; they can only bechanged by the optical fields. The lowest (single-exciton)manifold is accessible by linear optical techniques, such asabsorption spectroscopy and circular dichroism (CD),whereas the doubly excited (two-exciton) and higher mani-folds only show up in nonlinear spectroscopic techniques. Ahigh-level fluctuating excitonic Hamiltonian for polypeptidesis presented in Section 3. In Section 4 introduces the responsefunction approach for simulating the signals. The modeling ofthe coherent vibrational response involves the following keysteps:1. Protein and solvent configurations are generated by an

MD trajectory using existing molecular mechanics forcefields such as CHARMM,[58] GROMOS,[59] andAMBER.[60]

2. A fluctuating effective anharmonic vibrational–excitonHamiltonian, HS(t), and the transition dipole matrix m(t)for the relevant states, is constructed for each configu-ration. This must be a higher-level Hamiltonian, than themolecular mechanics force fields used in step (1) to modelthe structure.

3. Four-point correlation functions of transition dipoles arecalculated which describe the relevant motions andfluctuations.[61]

4. The response functions are calculated by specific combi-nations of the four-point correlation functions whichrepresent the quantum Liouville-space pathways relevantfor the chosen technique.[66]

Step 1 is well developed and documented and can use thebroad arsenal of available algorithms and software packages.We shall therefore focus on steps 2–4. Section 2 presents a

brief survey of the history of coherent multidimensionalspectroscopy. In Section 3 we review the methods forconstructing the fluctuating excitonic Hamiltonian for thepeptide amide bands. The theoretical framework for model-ing the nonlinear optical signal is introduced in Section 4,where we further introduce a simple exactly solvable Gaus-sian fluctuation model. In this sum-over-state (SOS)approach[62] the optical fields induce transitions betweensystem eigenstates, and the nonlinear response is attributed tothe anharmonicity of the system (note that harmonic vibra-tions are linear, their nonlinear response vanishes by inter-ference between quantum pathways). The SOS method ispractical for small peptides (as an example, the amide-I bandsof a peptide with less than 30 residues). The stochasticLiouville equations approach for describing chemicalexchange and spectral diffusion by incorporating externalcollective bath coordinates is introduced in Section 5. Appli-cations to the hydrogen bonding fluctuation dynamics inwater as observed in the OH stretch of HOD in D20 are givenin Section 6. A different method[48, 62, 63] more suitable for largebiomolecules such as globular proteins or membrane systemsis described in Section 7. In this approach the signal isconnected to the scattering of single excitations (quasiparti-cles) rather than transitions between states. The quasiparticleexpressions which scale more favorably with size can bederived by using equations of motion, the nonlinear excitonequations (NEE).[64] In Section 8 we demonstrate how aspecific choice of the signal wavevector can reveal doubleexcitations (double quantum coherence) which provide adifferent window for structure. Chirality-induced signals, 2Danalogues of circular dichroism, aimed at improving theresolution of 2DIR signals by exploiting the chirality ofpeptides, are presented in Section 9. Amyloid fibrils areaggregates formed by misfolded peptides associated withseveral human diseases, such as Alzheimer disease. Theirtoxicities strongly depend on their structures. 2DIR simula-tions described in Section 10 show great promise for retriev-ing structural information, not available by any othertechniques. A summary and future outlook of multidimen-sional techniques are presented in Section 11.

2. History of Multidimensional VibrationalSpectroscopy

IR absorption, provides a one dimensional (1D) projec-tion of molecular information onto a single frequency axis,through the linear polarization induced in the sample in thefield. Higher-order polarizations, and more complex molec-ular events, can be revealed by nonlinear spectroscopictechniques. Incoherent techniques,[65, 66] such as fluorescence,spontaneous Raman and pump–probe, do not depend on thephases of the laser pulses. Coherent techniques, in contrast, dogenerally depend on the phase. The modeling of nonlinearspectra is simplified considerably when the relaxation rates offrequency fluctuations (L) are either very fast or very slowcompared to their magnitude (D). They can then be incorpo-rated phenomenologically as homogeneous (D/L ! 1) orinhomogeneous (D/L @ 1) broadening, respectively. Picosec-




ond, electronically off-resonant, coherent anti-Stokes Ramanspectroscopy (CARS) measurements of vibrational dephasingperformed in the 1970s were believed to have the capacity todistinguish between the two broadening mechanisms,[67,68]

similar to the photon-echo technique.[69] By formulating theproblem in terms of multipoint correlation functions of theelectric polarizability, Loring and Mukamel[70] have shownthat this is not the case. The key lesson was that optical signalsshould be classified by their dimensionality, that is, thenumber of externally controlled time intervals rather than bythe nonlinear order in the field. Both photon-echo andelectronically off-resonant time resolved CARS signals arethird order in the external fields. However, the off-resonantCARS only has one control time variable t2 (see Figure 1).The other two times t1 and t3 are very short, as dictated by theHeisenberg uncertainty relation for an off-resonant transitionand carry no molecular information. The technique is thusone-dimensional, (1D) carries identical information to thespontaneous Raman and can not in principle distinguishbetween homogeneous and inhomogeneous line shapes.Based on this analysis, a 2D Raman analogue of the photonecho will require seven rather than three pulses. Suchexperiments have been carried out.[71–73] Closed expressionsderived for the multipoint correlation functions of a multi-level system whose frequencies undergo stochastic Gaussianfluctuations[74, 75] paved the way for the multidimensionalsimulations of such spectra.[76] Tanimura and Mukamel[77]

subsequently proposed a simpler, five-pulse, impulsive off-resonant Raman technique and showed how it can beinterpreted using 2D frequency–frequency correlation plots,in analogy with multidimensional NMR spectroscopy. Experi-ments performed on low-frequency (ca. 300 cm�1) intermo-lecular vibrations in liquid CS2

[78–87] were initially complicatedby cascading effects (sequences of third-order processes).These were eventually resolved.[88] Applications to liquidformamide were reported as well.[89] The same idea was thenproposed[48, 65,90] and carried out for vibrational transitions inthe IR[57, 91–93] and for electronic spectroscopy in the visibleregion.[94–97] IR techniques require fewer pulses than Raman,since each transition involves a single field, rather than two.The necessary control of the phase of some or all laser pulses,which is straightforward for radiowaves (NMR) is consider-ably more challenging for higher frequencies.

The first frequency–frequency 2D IR measurement wascarried out by Hamm and Hochstrasser,[98] who employed apump–probe technique with two IR pulses with a narrow (ca.10 cm�1) pump and a broad (130 cm�1) probe pulse. The signalfield was dispersed in a spectrometer and recorded against thepump and the dispersed signal frequencies. This studydemonstrated how the cross-peaks can be used to investigatethe structures of small peptides.[99] The diagonal peaks revealthe energies of the localized carbonyl C=O vibrational mode,while the cross-peaks are directly related to the couplingsbetween the modes which depend on the peptide structure.The cross-peak intensities and anisotropies of a cyclic rigidpenta-peptide were connected to the 3D structure with thehelp of a simple model for the coupling.

A quantitative analysis requires higher level modelHamiltonian operators, which were developed for small

peptides. The central backbone structure of trialanine inaqueous solution was investigated, using polarization sensi-tive two-dimensional (2D) vibrational spectroscopy on theamide I mode.[100] The stimulated IR photon echo of N-methylacetamide (NMA),[101] a molecular mimic of a singleamide unit, was then measured and used to determine thevibrational frequency correlation function. These results areoften used as a benchmark for effective Hamiltonian oper-ators. Experiments performed on polypeptides are widelyused for benchmarking the couplings and transition dipoles.2DIR spectra of a series of doubly isotopically substituted 25-residue a-helices were reported in 2004. 13C and 18O labelingof at known residues on the helix permitted the vibrationalcouplings between different amide I modes separated by one,two, and three residues to be measured.[102] Two similarstudies of a beta hairpin,[56] a 310 helix and another type of a-helical peptide[103] were reported. Other small moleculesincluding DNA[104] and a rotaxane (molecular ratchet)[105]

have been studied.The successful applications of 2DIR to small peptides

triggered the investigation of larger systems. Tokmakoff et al.identified a characteristic “Z” shape photon-echo spectrum ofthe b-sheet motif proteins by comparing several globularproteins with increasing b-sheet content.[57] a-Helices showeda flattered “figure-8” line shape, and random coils gave rise tounstructured diagonally elongated bands.[106] Righini and co-workers studied the local structure of lipid molecules indimyristoylphosphatidylcholine (DMPC) membranes usingisotopic labeling of the carbonyl moieties in the mem-branes.[107, 108] In a 2D line shapes study of the amide-I bands(backbone carbonyl stretch) for 11 residues along the lengthof a transmembrane peptide bundle, Zanni et al. measuredthe homogeneous and inhomogeneous widths of vibrationalmodes that reflect the structural distributions and picoseconddynamics of the peptides and their environment.[93] 2DIRstudies of misfolded peptide aggregates (amyloid fibrils) werereported.[109–111]

Time-resolved measurements can monitor the pico- tonanosecond dynamics by following the variation of the cross-peaks with time,[112] Hamm and co-workers had monitored theunfolding of a tetra-peptide triggered by breaking thedisulfide bridge between the first and the third residues by aUV pulse.[22, 92] The cross-peaks reveal the hydrogen-bondingdynamics on a timescale of a few picoseconds. Based on 2DIRof the amide I vibrations, Tokmakoff reported the steady-state and transient conformational changes in the thermalunfolding of ubiquitin.[113] Equilibrium measurements areconsistent with a simple two-state unfolding. However, thetransient experiments show a complex relaxation pattern thatvaries with the spectral component and spans six order ofmagnitudes in time. Using time-resolved IR spectroscopy,Hamm et al. could report strongly temperature-dependentnon-exponential spectral kinetics of the folding and unfoldingof a photoswitchable 16-residue alanine-based a-helicalpeptide from a few picoseconds to almost 40 ms over thetemperature range 279–318 K.[114] Both processes show acomplex stretched-exponential response, indicating a broaddistribution of rates. Environment effects on the vibrationaldynamics of tungsten hexacarbonyl in cryogenic matrices


Chemie



were investigated using an infrared-free electron laser bymeasuring the population relaxation time T1 in pump–probeand the dephasing time T2 in a two-pulse photon echo.[115] Fast(less than a few ps) enzyme dynamics at the active site offormate dehydrogenase (FDH) in complex with azide (N3, ananomolar inhibitor, and a transition state analogue) andnicotinamide (NAD + ) were detected by IR photon-echomeasurements. These studies show that the active site of thereactive enzyme complex near the catalytic transition stateexhibits the fast dynamics required to explain the kinetics ofseveral enzymes.[116] Harris and co-authors showed that 2DIRspectroscopy can provide direct information about thetransition-state geometry, time scale, and reaction mecha-nisms by tracking the transformation of vibrational modes as[Fe(CO)5] crossed a transition state of the fluxional rear-rangement.[117]

Ultrafast IR–Raman spectroscopy (mid-IR pump andRaman probe pulse) were applied to study fast energy-transfer dynamics in liquid water, HOD in D2O, andmethanol.[118, 119] The hydrogen-bonding structure and dynam-ics in liquid water have been extensively studied by 2DIR.Tokmakoff and co-workers investigated rearrangements ofthe hydrogen-bond network by measuring fluctuations in theOH-stretching frequency of HOD in liquid D2O. Thefrequency fluctuations were related to intermolecular dynam-ics. The model reveals that OH frequency shifts arise fromchanges in the electric field acting on the proton. At shorttimes, vibrational dephasing reflects an under-damped oscil-lation of the hydrogen bond with a period of 170 femto-seconds. At longer times, vibrational correlations decay on a1.2 picosecond timescale as a result of collective structuralreorganizations.[120] A combined femtosecond 2DIR andmolecular dynamics simulations study focused on the stabilityof non-hydrogen bonded species in an isotopically dilutemixture of HOD in D2O. Hydrogen-bonded configurationsand non-hydrogen-bonded configurations were shown toundergo qualitatively different relaxation dynamics.[121]

Molecular dynamics electronic-structure calculations wereused to obtain the time-correlation functions (TCF) for twowater force fields, TIP4P and SPC/E.[122] The TCFs serve asinputs to simulations of 2DIR spectra. Comparison withexperiment demonstrates that both force fields overempha-size the fast (300–400 fs) fluctuations and do not account forthe slowest fluctuations (1.8 ps). The vibrational echo corre-lation spectra provide a good test for the TCF. Temperaturedependence of the OH stretch photon-echo signal of liquidH2O showed that the frequency (thus structural) correlationsdecrease from 50 fs to 200 fs as temperature decreases from297 to 274 K, which suggests a reduction in dephasing bylibrational excitations.[123] Simple anions (CN� , N3

�) havebeen used as probes of the fluctuations of water hydrogen-bonding networks.[124–126]

Electrostatic interactions are crucial for enzyme activityand drug design. For example, the noncovalent electrostaticcouplings of cofactors are sufficiently weak to allow forreversible binding.[127, 128] 2DIR should provide a direct meansfor monitoring the electrostatic environment and its fluctua-tions. Artificial chromophores, such as nitriles, could beinserted in specific sites in the active region.[129] A 2DIR study

of an HIV drug molecule containing two nitrile groups[130]

complexing to the reverse transcriptase of HIV-1 showsspectral splitting attributed to the binding environment.Several bond types, including nitriles, carbonyls, carbon–fluorine, carbon–deuterium, azide, and nitro bonds were usedas probes for electric fields in proteins using vibrational Starkspectroscopy. The measured Stark shifts, peak positions, andabsorption cross sections may be used to design amino acidanalogues or labels to act as probes of local environments inproteins.[131] 2D techniques[118,133] can monitor vibrationalenergy transfer pathways.[132] Vibrational energy relaxationrates were simulated by employing the semiclassical approx-imation of quantum mechanical force–force correlationfunctions.[134]

2D techniques may also be used to study reaction rates,mechanisms, and yields. Small peptides at thermal equilibri-um in solution rapidly (within 10–100 ps) swap amongdifferent configurations. The dynamics of these transientspecies can influence the folding. Hochstrasser and Kim [135]

and Fayer et al.[136] independently carried out a 2DIRanalogue of chemical exchange for the investigation ofultrafast hydrogen-bonding dynamics of solute–solvent com-plexes. Hamm et al. employed non-equilibrium 2DIRexchange spectroscopy to map light-triggered protein ligandmigration.[137] Bond connectivity in molecules has beenmeasured based on relaxation-assisted 2DIR signals, for thistwo parameters are decisive, a characteristic intermodeenergy transport arrival time and a cross-peak amplificationcoefficient. 2DIR spectra of the coupled carbonyl stretchingvibrations of [Rh(CO)2(C5H7O2)] in hexane detected withspectral interferometry characterizes the structure with a20 ps time window.[138]

3. Hamiltonian Operators for the Amide Vibrationsof Polypeptides

Vibrational spectra are commonly described by normalmodes, which represent the collective motions of the atomswhen all anharmonicities are neglected. The normal-modefrequencies and individual atom displacements may becalculated from molecular-mechanics force fields imple-mented in standard MD codes. These are parameterized torepresent slow backbone motions. High-frequency vibrationssuch as the amide bands of peptides require more accurateab initio calculations.

A peptide can be viewed as a chain of beads connected byamide bonds (O=C-N-H; Figure 3). These have a partialdouble-bond character and because of steric effects arealmost exclusively in the trans configuration. The areabetween two consecutive a-carbon atoms (peptide unit) isthus rigid and planar. The peptide backbone structure isdescribed by two dihedral Ramachandran angles f and y peramide bond. The IR spectrum of the backbone peptide bondsconsists of four amide vibrational bands, known as theamide I, II, III and A.[139–41] These amide bands originatefrom the coupled localized amide vibrations on each peptideunit (local amide modes (LAMs)) The localization may beviewed by expanding the molecular charge density 1(r) in




nuclear displacements [Eq. (1)] where qmi is the i’th vibra-tional mode of the m’th peptide bond.

The transition charge density (TCD) @1/@qmi[142] repre-

sents the electronic structure change induced by the qmi

vibration. The 0.01 esu/Bohr (TCD) contours of the foramide vibration bands of N-methyl acetamide (NMA), amodel system of the amide bond, are shown in Figure 4. The

amide III(ca. 1200 cm�1), II(ca. 1500 cm�1) modes are attrib-uted to bending motion of the N�H coupled to the C�Nstretching. The 1600–1700 cm�1 amide I mode originates fromthe stretching motion of the C=O stretch coupled to in-phaseN�H bending and C�H stretching. The amide A (ca.3500 cm�1) is almost purely the N�H bond stretch.[139, 140,143]

All the TCDs are highly localized on the four atoms (O, C, N,and H) forming the amide bond. The overlap of the amideexcitations between different amide bonds is small and islimited to nearest-neighbor peptide bonds. By parameterizingthe Hamiltonian and transition dipole elements of all theamide bands (I, II, III and A) by the Ramachandran angles,repeated electronic structure normal mode calculations canbe avoided for various conformations.

The sensitivity of the amide vibrational transitions to thelocal structure and the hydrogen-bonding environment makesthem ideal candidates for distinguishing between varioussecondary structural motifs and monitoring effects of thechanging environments.[144] The intense and spectrally iso-lated amid-I band is particularly suitable for structuredetermination. Its frequency variation with the secondarystructure and conformation is widely used as a marker inpolypeptide- and protein-structure determina-

tion.[56, 57,105, 143–146] a-Helical peptides have amide-I bandsbetween 1650–1655 cm�1. b-Sheets usually have a strongband between 1612–1640 cm�1 and a weaker band at approx-imately 1685 cm�1. Random structures generally have a1645 cm�1 band, which is close to the frequencies associatedwith a-helix. The antiparallel b-sheet structure shows a strongamide-II band between 1510 and 1530 cm�1, whereas aparallel b-sheet structure has higher frequency bands (1530–1550 cm�1). Deuterium substitution results in a substantialshift to lower frequency (to ca. 1460 cm�1). The amide III IRband is typically weaker than the amide I and II. Deuterationalso shifts the amide-III band to lower frequencies (960–1000 cm�1). This band is usually not correlated with proteinsecondary structure, but is sensitive to hydrogen bonding andlocal Ramachandran angles.[19] The amide-III band is some-times used in combination with the amide-I band to distin-guish the b-sheet and disordered structure which is notgenerally possible with only the amide-I band. The overlap ofthe amide A with the intense O�H band of water complicatesits observation and interpretations.

A vibrational Hamiltonian operator adequate for 2DIRsimulations of peptides has been constructed by expandingthe potential to the 6th order in LAMs within each peptideunit, to the 4th order for neighboring couplings, and to the 2ndorder for non-neighboring electrostatic couplings [Equa-tion (3) of ref. [148] and Eq. (2)].[142] Interactions betweenLAMs with non-overlapping TCD are purely electrostaticand are given by Equation (2).

By diagonalizing the local Hamiltonian for each amidebond, 14 local amide eigenstates (four fundamentals, fourovertones, and six combinations) are obtained in the energyrange 0–7000 cm�1 for a single peptide unit, which will bedenoted as local amide states (LAS). We define the excitoncreation and annihilation operators for a’th LAS on the m’thunit Bma� jmaihm0 j and BByma� jm0ihma j where jm0i is theground state. These satisfy the Pauli commutation relations:[Bma,B

ynb] = dn,mda,b(1��cB

ymcBmc)�dm,nB

ymbBma.

The peptide Hamiltonian is then recast in terms of theseoperators [Eq. (3)].[148]

The first term represents the local Hamiltonian. Thecouplings between neighboring peptide units (the secondterm) were computed as a function of the Ramachandran

Figure 3. The amide bonds and Ramachandran angles (f and y). Theplanes marked by gray lines are peptide units. “sc” represents sidechains.

Figure 4. Transition charge densities (TCD) for the four amide modesof N-metylacetamide (NMA). The 0.01 esu/Bohr contour is shown.Violet and brown contours represent positive and negative values,respectively.


Chemie



angles (f and y) at the BPW91/6-31G(d,p) level of DFTusinga 4th order anharmonic vibrational potential of variousglycine dipeptide (GLDP) configurations. An electrostaticmodel is used for couplings between non-neighboring units(the last two terms). The transition charge density couplings(TCDC) was expanded in terms of multipoles to the 4th rank.This treatment results in terms for dipole–dipole (ca. R�3),dipole–quadrupole (ca. R�4), and quadrupole–quadrupoleand dipole–octupole (ca. R�5) interactions where R is thedistance between units.

Torii and Tasumi (TT) constructed such a map[146] for theamide I neighboring coupling using restricted Hartree–Fock(RHF) electronic structure calculations of glycine dipeptide(GLDP). The amide I through-space coupling between thenon-neighboring peptide units was approximated by thetransition dipole coupling model (TDC).[149, 150] The magni-tude, direction, and location of the transition moment werefitted to reproduce the ab initio coupling constants betweenthe second nearest amide units (the magnitude of thetransition dipole was (@m/@q) = 2.73 DA�1 with 10.08 angleto the C=O bond). Gorbunov, Kosov, and Stock[151] derived asimilar potential map at higher levels (MP2 and B3LYP).Woutersen and Hamm approximated through-space TCDCwith Mulliken partial charges of a DFT calculation to includehigher multipole contributions for the amide I vibration ofmethylacetamide (NMA),[152] which was later improved byusing multipole-derived charges.[153] The accuracy of theamide I local frequencies and IR intensities with respect toreference DFT calculations was improved (0.1 cm�1 infrequency and 0.02 in IR intensity correlation) by includinghigher multipoles (see Table V of Ref. [153]). Transitionmultipole couplings extend the transition dipole couplingsto include higher multipoles of all the amide modes. Thehigher multipole contributions are more important for amide-II, -III, and -A modes than I since the amide II and III aremore delocalized over the peptide bond, and the amide A hasa smaller transition dipole (see Figure 8 of Ref. [142]).

The dipole coupling with the radiation field is given by[Eq. (4)].

To account for chirality this approach was extended toinclude magnetic moments. Derivatives of magnetic momentswith respect to the LAM depend on the Ramachandranangles y and f. A magnetic moment derivatives map wasobtained by DFT calculations of a chiral model peptide unitwhich has a similar structure to NMA (see Figure 1 ofRef. [148]). These were calculated based on the atomic axialtensor and the normal modes.[154]

3.1. Electrostatic Fluctuations of the Local HamiltonianOperators

The local Hamiltonian operator [Eq. (3)] depends on theelectrostatic environment induced by the surrounding peptideresidues and the solvent. For example, the amide I frequen-

cies are shifted to the red by hydrogen bonding with water.Electrostatic modeling of the fluctuating local Hamiltonianrequires repeated ab initio vibrational potential calculationsof the peptide bonds surrounded by the partial charges of thesurrounding peptide residues and the solvent. Simulation of2DIR line shapes in NMA require the construction of aHamiltonian along the MD trajectory for which there aretypically around 105 snapshots. These repeated ab initiocalculations can be avoided by an electrostatic parameter-ization of the Hamiltonian operator. Linear Stark modelingwhich uses the electric field at some reference point[155] worksfor smaller chromophores.[120, 156, 157] This model is not ade-quate for the amide vibrations of proteins where the non-uniform electric field across the peptide bond should be takeninto account.[142]

Ham and Cho (HC) obtained a map which parameterizesthe amide I frequencies as a linear function of the electro-static potentials at the C, O, N, and H and two methylsites[158,159] by a least-square fit of the normal mode frequen-cies of NMA–water clusters at the RHF level. Schmidt,Corcelli, and Skinner[160] constructed a similar map (SCS) ofthe deuterated NMA (NMAD) amide I frequency, where thefrequency was parameterized as a linear function of theelectric fields at the C, O, N, and H atoms, and the electronicstructure calculations were made at the DFT level. Watsonand Hirst found that the accuracy of NMA amide I frequen-cies in water is improved by sampling the electrostaticpotential at additional points in the amide bond (mid pointsof CO, CN, and NH).[161]

We have parameterized the fundamental, the overtone,the combination frequencies, and transition dipoles of all theamide modes (III, II, I, and A) as a quadratic function of themultipole electric field up to the 2nd derivatives of the electricfield at a midpoint of amide oxygen and hydrogen atoms(HM map).[142] The map was constructed by eigenstatecalculations of the 6th order anharmonic DFT (BPW91/6-31G(d,p)) vibrational potential in five relevant normal modesof NMA in the presence of different non-uniform multipoleelectric fields. The fundamental frequencies as well asanharmonicities are parameterized, and geometry changesand mode mixing induced by the multipole electric field areincluded. The average of and the correlations between thefundamental and anharmonicity frequency fluctuations deter-mine the relative positions and intensities of two positive(stimulated emission/ground state bleach) and negative(excited state absorption) peaks of the nonlinear IR sig-nals.[162] Unlike the other map, this map does not involve afitting to a specific solvent. A similar approach was lateradopted for the amide I frequency by Jansen and Knoester[163]

who constructed the amide I single mode anharmonic vibra-tional potentials for NMA embedded in a set of differentsolvent charge distributions. The amide I frequencies wereparameterized by the electric field and gradients at the C, O,N, and H atoms. The map does not include mode mixing.Frequencies and IR intensities of a pentapeptide in severalgas-phase configurations[161] calculated by this map combinedwith transition charge couplings and neighboring couplingmap were in good agreement with DFT calculations,[153]

similar to the HC map.




The Torii and Tasumi couplings were used with our localHamiltonian operator in earlier applications.[164] The fullHamiltonian [Eq. (3)] was employed for two a helical pep-tides (SPE3[145]) reported later in this Section.

The segment made of a given amide residue and twoneighboring neutral groups of the CHARMM27 forcefield[165] was used as the basic chromophores in the electro-static interaction calculations(Figure 3). The effect of the restof the protein and the solvent is described by a fluctuatingelectrostatic field. The electrostatic potential U is expanded tocubic order in local Cartesian coordinates Xa. (a, b = x, y, z)around the midpoint between amide oxygen and hydrogenatoms of the amide bond [Figure 5) and Eq. (5)].

Apart from the reference U0, Equation (5) has19 independent parameters arranged in a vectorC = (Ex,Ey,Ez,Exx,Eyy,Ezz,Exy,Exz,Eyz,Exxx,Eyyy,Ezzz,Exyy,Exxy,Exxz,Exzz,Eyzz,Eyyz,Exyz) (note the symmetry Eab = Eba).

The components of C are determined at each time step bya least-square fit to the electric field which is sampled at 67points in space spanning the TCD region of the four amidemodes (Figure 5). We expect the electrostatic potential in theregion of large TCD to affect the IR activity of that vibration.Four sampling points at C, O, N, and H atom positions werenot sufficient to predict the solvent frequency shits (especiallyfor the amide-II and -III bands).[142] This finding is consistentwith the result of Watson and Hirst who found that increasingthe number of sampling points improved the accuracy of theamide I frequency.[161] The parametric dependence of theanharmonic force field on the electrostatic multipole coef-ficients C was obtained for NMA[142] at the BPW91/6-31G(d,p) level.[166] This functional is known to give accurateamide vibrational normal mode frequencies wma(0) of pep-tides.[167] Analytic energy gradient in the presence of multi-

pole field was implemented in the Gaussian 03 code[168] tocompute the higher derivatives.[169,170]

The LAS were calculated for a grid of electrostaticmultipole coefficients C by diagonalizing the local Hamilto-nian operator expanded in a harmonic basis set. TheARNOLDI matrix diagonalization algorithm was employedin these vibrational configuration interactions (vibrationalCI) calculations. The vibrational transition frequency fromthe ground state to LAS a and the transition dipole momentsbetween LAS a and b at the m’th peptide unit were expandedto quadratic order in C [Eq. (6) and (7)]. The gas-phasefrequencies were taken from experiment[171] and Oð1Þa and Mð1Þ

ab

are 19-component vectors representing the first derivative ofthe frequency and the transition dipole with respect to the C.Oð2Þa and Mað2Þ

ab are the second derivative 19 � 19 matrices.

To trace the origin of the electrostatic effects on the amidefrequency shifts, the C=O and N�H bond lengths obtained byenergy minimization for the various field values were para-meterized in terms of C.[142] Strong correlations are seen in thescatter plots of the four amide fundamental frequencies withC=O and N�H bond length (Figure 6). These suggest thatstructural changes of NMA caused by the electricfield[142,146, 155, 172,173] are responsible for the frequency shifts.The positive correlations of the two bending frequencies withthe N�H bond length are ascribed to the fact that thehydrogen bonding to H6 causes a longer N�H bond length

Figure 5. Left: NMA molecular structure and coordinate system usedfor the anharmonic force field and the electrostatic potential. The fouramide atoms (O4, C3, N2, and H6) are in the x,y plane. The origin isthe middle point of the oxygen (O4) and hydrogen (H6) atoms. Right:Contour plots of the non-uniform electric field (Ex and Ey) of NMA inH2O. Red circles represent the four amide atoms (O4, C3, N2, andH6). The sampling points are shown by the blue crosses.

Figure 6. Scatter plots of amide frequencies versus bond lengths.Linear fits are w = 6549.2�3905RCO (amide-I versus C=O bondlength), w = 24259�20426RNH (amide-A versus N�H bond length),w =�3066+ 4278RNH (amide-III versus N�H bond length) andw = 2768+ 4204RNH (amide-II versus N�H bond length). The gas-phase values are marked by crosses.


Chemie



and stiffens the potential more along the amide II and IIIbending modes by stabilizing the parallel N2�H6···OH2

structure.The simulated amide I solvent peak shifts (�59 cm�1) and

line widths (29 cm�1) of NMA in water are in good agreementwith experiment (�80 cm�1 and 29 cm�1 respectively). Thiseffective Hamiltonian operator was applied to SPE3, a 16-residue a-helical peptide (YGSPEAAA(KAAAA)3r, r rep-resent d-Arg).[145] The fluctuating Hamiltonian operator wasconstructed for 100 snapshots obtained from a 2 ns MDtrajectory.[164] The vibrational eigenstates were calculated bydiagonalizing the HM Hamiltonian operator. A good mea-sure of the coherence length Ln of the n’th vibrationaleigenstate is provided by the participation ratio[Eq. (8)][54, 65,174] where Cv,ma is the expansion coefficient ofthe n’th eigenvector on LAS a at the n’th peptide unit.

The distributions of Ln over the frequencies of theeigenstates in four amide fundamental regions are shown inFigure 7. In the amide I region, the lower frequency eigen-

states (ca. 1600 cm�1) are mostly localized on one amide bond(hLni� 1), the higher frequency eigenstates are delocalized.The higher frequency amide III eigenstates (ca. 1300 cm�1)are localized. In the amide II region, there are two or threepeaks in participation ratio distribution. The amide II funda-mentals are the most delocalized with hLni= 2.3, owing to thelarger neighboring couplings and transition moments, andsmaller diagonal frequency fluctuations. The amide-III and -Ifundamentals are delocalized over 1.6 and 1.8 amide bonds.The amide A modes are highly localized (hLni= 1.0) owing tothe small transition moment and large frequency fluctuations.

This localization is good news for the interpretation of 2DIRsignals in terms of local structure.

4. Liouville-Space Pathways for Coupled LocalizedVibrations

Coherent optical signals can be classified according totheir power-law dependence on the driving field intensities.[66]

The signals are related to the polarization, P(t), induced bythe external electric fields. The induced-polarization can beobtained perturbatively by expanding the density matrix 1(t)in powers of the external fields.[66] The third-order responsefunction, Rð3Þn4n3n2n1

(t3,t2,t1) represents the lowest order contri-bution to the induced polarization in isotropic systems[Eq. (9)] where t1,t2,t3 represent the interaction time intervalsbetween successive interactions with the optical pulses, E(r,t ;see Figure 1); nj are the Cartesian components of the fieldsand polarizations. The response functions are system propertytensors that contain all relevant molecular information. R(1) isa second-rank tensor connecting two vectors (E and P).Similarly, R(3) is a fourth-rank tensor.

A heterodyne-detected four-wave mixing experiment(Figure 1)[175] involves four pulses. In ideal impulsive meas-urements the pulses are temporally ordered, well separated,and much shorter than the relevant molecular timescales.Under these conditions, all integrations in Equation (9) canbe eliminated and the optical signal is simply proportional tothe response function itself. The third-order response isillustrated in Figure 8. The system is initially in thermal

equilibrium, and the Green�s function g(tn) describes the freemolecular time evolution (without the fields). At time 0 itinteracts with the first pulse (vn1

), propagates freely during t1

(g(t1)), interacts with second pulse (vn2) at t1, propagates

during time t2 (g(t2)),interacts with third pulse (vn3) at t1 + t2,

propagates during t3 (g(t3)), and finally interacts with thesignal mode (vn4

) at t1 + t2 + t3 to create the response. Thedipole operator can act three times either on the ket or the bravector.

The third-order response function is thus given by a sumof 23 = 8 four-point correlation functions[66] which constitutethe eight basic Liouville space pathways [Eq. (10)].

Figure 7. Distribution of the participation ratio (PR) versus frequencyin the amide-III, -II, -I, and -A regions. Average PRs are 1.6 (amide III),2.3 (amide II), 1.8 (amide I), and 1.0 (amide A).

Figure 8. The schematic representation of third-order response func-tion.




Different techniques can select some of the possible termsin Figure 8, depending on the pulse configuration and thedetection mode. To compute the signals, the electric field Emust be expanded in modes [Eq. (11)].

The pulse j = 1, 2, 3 and s is centered at tj, with wavevectorkj, carrier frequency wj, phase fj and complex envelopeenj

(t�tj). The three incoming pulses are labeled 1,2,3 and thesignal as s. The k1 pulse comes first, followed sequentially byk2, k3, and ks. The heterodyne signal S(t), defined as thechange in the transmitted intensity of mode s induced by theother three beams (1–3), is related to P(3), the third orderpolarization, and the external fields [Eq. (12)] where the r isintegrated over the interaction volume in the sample.

Coherent nonlinear signals are highly directional and areonly generated when ks lies along one of the phase-matchingdirections: ks =� k3� k2� k1 (with the corresponding fre-quencies ws =�w3�w2�w1). This important feature ofcoherent spectroscopy stems from the fact that we add thefield amplitudes generated by different molecules, when thesample is much larger than the optical wavelength.[297]

Random phases then cancel the signals in other directions.Incoherent signals, such as fluorescence are obtained byadding the intensities (amplitude squares), and the signals areessentially isotropic. A whole host of names and acronymshave been used for various combinations of vectors and timeintervals (e.g. photon echo, transient grating, CARS,HORSES, etc). NMR spectroscopy has it own set ofacronyms. We shall avoid this nomenclature and simplyclassify the signals into four basic techniques: kI =�k1 + k2 +

k3, kII = k1�k2 + k3, kIII = k1 + k2�k3, and kIV = k1 + k2 + k3.The dominant contributions to resonant signals only come

from terms obtained when the field and molecular frequen-cies in Equation (9) have an opposite sign. Other (same-sign)highly oscillatory terms may be safely neglected. Using thisrotating wave approximation (RWA), each phase-matchingsignal is described by a specific combination of Liouville-

space pathways. As an example, forthe kI technique we get Equa-tion (13).

The dependence ofRð3Þ

ks, n4 n3 n2 n1

(t3,t2,t1) on the wavevectorcomes by selecting the RWA path-ways.

To invoke the RWA the molec-ular model must be specified. Theamide band energy level schemeconsists of three well-separatedbands (Figure 9). Only transitionsbetween the ground state, g, andthe first excited states manifold, e,and between the first and secondexcited state manifold, f, are allowed.The response functions may be cal-culated by summing over all possibletransitions among vibrational eigen-states. The nonlinear response van-ishes for harmonic vibrations and may thus be attributed tothe anharmonicities. The terms that contribute to the signalcan be represented using Feynman diagrams which show theevolution of the density matrix. These are constructed usingthe following rules:1. The density matrix is represented by two vertical-lines

which represent the ket (left line) and the bra (right line)vectors.

2. Time runs vertically from bottom to top.3. Each interaction with the radiation field is represented by

a wavy line. An arrow pointing to the right and labeled kj

represents a contribution of ejexp(�iwjt + ikj·r) to thepolarization. An arrow pointing to the left represents acontribution of the term e*

jexp(iwjt�ikj·r) wj(> 0).

4. Each diagram has an overall sign of (�1)n where n is thenumber of interactions from the right side (bra) (aninteraction v that acts from the right in a commutator inthe Liouville equation carries a minus sign).

The Feynman diagrams for the kI technique depicted inFigure 10 show the state of the density matrix during eachtime interval. Computing the signals generally involvesmultiple integrations over the pulse envelopes [Eq. (9)].

Figure 9. Energy-levelscheme for the systemsconsidered. g is theground state, e is thefirst excited-state mani-fold, and f is thesecond excited-statemanifold. The transi-tions that can beinduce by the pulsesare shown as mge andmef.

Figure 10. Double-sided Feynman diagrams representing the Liouvillespace pathways contributing to the kI signal in the rotating-waveapproximation. The three pathways are known as excited-state emis-sion (ESE), the ground-state bleaching (GSB), and excited-stateabsorption (ESA).


Chemie



Clearly, the shape and relative phases of the pulses areimportant factors which affect the signal. Coherent controland pulse shaping algorithms may be used to design signalsthat meet desired targets.[64] Herein we focus on ideal time-domain techniques where the pulses are well separatedtemporally. Multidimensional signals are displayed in thefrequency domain by performing the multiple Fourier trans-form of Sð3Þ

ks

(t3,t2,t1) with respect to the time intervals between

the pulses. We shall consider the signal given in Equa-tion (14).

This signal is given by ref. [176] in the form of Equa-tions (15)–(17).

These expressions show how the pulse envelopes selectthe transitions lying within the pulse bandwidths. The terms eand e’ run over the first excited state manifold and f includesthe second excited states manifold (Figure 9). The terms w1,w2, and w3 are the carrier frequencies of the first three pulses.wab = (ea�eb)/�h are the transition frequencies where the es arethe state energies and xab = wab�igab are complex transitionfrequencies which include the dephasing rates g. In theimpulsive (broad bandwidth) limit we simply set e(w) = 1.

4.1. Simulating 2DIR Spectra of Small Peptides with GaussianFrequency Fluctuations

We now turn to a special class of fluctuation models[55,62]

which may be solved exactly, yielding compact, closedexpressions for the response functions. These have beenapplied for modeling 2DIR signals of small peptides with lessthan 30 residues.

We assume purely diagonal (energy) fluctuations thatconform to Gaussian statistics. The fluctuations are smallcompared to energy-level spacings. This is the case when the

energies are modulated by collective coordinates expressed assums of harmonic coordinates. However, the model holdsmore broadly, thanks to the central limit theorem, when thecollective coordinates are given by sums of many bathcoordinates, each making a small contribution. One notableexample is the Marcus theory of electron transfer[177] wherethe collective coordinate is the electric field at a given site,given by the sum of contributions from all surroundingcharges in the solvent. Using the Condon approximation wecan neglect fluctuations of the magnitude of the transitiondipole.

The response functions can be calculated exactly using thesecond-order Cumulant expansion. To that end, we defineUma(t)�wma(t)�wma representing the fluctuations of thetransition frequencies, where wma is the average transition

frequency. The two-time correlation func-tion of U is given by Equation (18) whereC’(t) and C’’(t) are the real and imaginaryparts of C and t12 = t1�t2. We furtherdefine the line-broadening functions asEquation (19): Using the fluctuation-dis-sipation relation between C’ and C’’, gmn(t)can be expressed as Equation (20) wherethe C’’nm(w) known as the spectral densitygiven by Equation (21). The real and theimaginary parts of gnm(t) are responsiblefor line-broadening and spectral shift,respectively. The third-order nonlinearresponse functions can be expressed interms of g(t).[66] We denote this model asthe Cumulant expansion of Gaussian fluc-tuations (CGF).

Simulated kI and kIII signals of all the amide modes of N-methyl acetamide (NMA) in water in the cross-peaks regionsare shown in Figure 11. The simulations reproduce theamide I and II anharmonicities obtained by the recent cross-peak experiment (calcd: 14 cm�1 and 13 cm�1; exp: 12 cm�1

and 10 cm�1, respectively).




The frequency–frequency correlation function of twovibrational transitions [Eq. (18)] can be represented asEquation (22) where Dmm�

ffiffiffiffiffiffiffiffiffiffiffiffiffiffi

hU2mai

p

is a fluctuation ampli-tude, Cmn is a normalized correlation function (Cmn(0) = 1) ,and hmn�hUmaUnai/

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

hU2maihU2

naiq

is the correlation coefficientwhich varies between 1 (full correlation) through 0 (nocorrelation) to �1 (anti-correlation).

To investigate the sensitivity of coherent IR signals tocorrelated frequency fluctuations, the amide I–III photon-echo cross-peak of NMA, Im[SkI

(W1,t2 = 0,W3)] are shown inFigure 12 for various combinations of the correlationscoefficients hmn. Negative and positive peaks of the signalcorrespond to the ESA and ESE/GSB pathways of Figure 10respectively. Correlations between the amide I and III (h13)contribute to the negative components, and correlationbetween the amide III and the combination state I + III(h19) contributes to the positive component. The negativepeak becomes weaker and broader as h13 is varied from + 1 to�1, but does not depend significantly on h19. The positive

peak becomes smaller and broader elongated more in W3

direction as h19 goes from full correlation (+ 1) to anti-correlation (�1). The actual simulated values (anti-correlatedh1,3 =�0.71 and correlated h1,9 = 0.63) gives weaker signalsthan in the fully correlated case.

For systems with several local minima whose dynamicscan be divided into well-separated time regimes, the inhomo-geneous CGF method can be adopted.[76] The spectrum isobtained by summing over contributions of the slowlyinterconverting configurations, each represented by theCGF. This method is illustrated for the simulation of theamide I region of a specific tryptophan zipper peptide, trpzip2in a b-hairpin conformation and its 13C isotopomers. b-Hairpins are common protein structural elements whichprovide an important model system for the folding kinetics oflarger proteins. Their structure and folding dynamics havebeen studied extensively. One structural motif, the tryptophanzipper (trpzip), greatly stabilizes the b-hairpin conformationin short peptides (12 or 16 � in length). Trpzips are thesmallest peptides to adopt a unique tertiary fold withoutrequiring metal binding, unusual amino acids, or disulfidecrosslinks. 500 snapshots with 2 ps time intervals wereselected from the 1 ns trajectory and used for the inhomoge-neous averaging. A 5.5 cm�1 homogeneous dephasing rate[178]

g was added. Two electrostatic maps (HC[158] and HM[142] asdescribed in Section 3) were used to compute the local modefrequencies in solution. Spectra of unlabeled samples (UL)and those with 13C isotope labeling at specific residues arecalculated. The sample with a b strand residue (the secondresidue) labeled is denoted L2, while the sample with a turnresidue (the seventh) labeled is denoted L7. The simulatedsignals are compared with experiment in Figure 13 andFigure 14.

The experimental absorption spectra of trpzip2 13Cisotopomers in the amide-I region are displayed inFigure 13. They show two main transitions: the stronger lowfrequency transition, around 1640 cm�1, is due to inter-chainin-phase C=O motions and intra-chain out-of-phase C=Omotions, whereas the weaker high-frequency transition,around 1675 cm�1, is mainly due to the inter-chain out-of-phase C=O motions and intra-chain in-phase C=O motions.The two isotopomers show different 13C effects: the 13C bandis shifted 10 cm�1 to the red in L7 than in L2 (1590 vs.1600 cm�1). Simulated spectra are shown in Figure 13 panel B

(HC Hamiltonian operator) and Figure 13 panel C (HMHamiltonian operator). The high-frequency componentis slightly stronger for the HM, but overall both modelsreproduce the main spectral feature of the unlabeled b-hairpin. In addition, both predict a small difference inthe observed 13 C-shifts between L2 and L7. Thedifference is slightly larger in the HM simulation.

The top row in Figure 14 shows the experimentalkI + kII spectra of the trpzip2 13C isotopomers. InFigure 14 panel A, the diagonal signals are due to 0-1 (red) and 1-2 transitions (blue). The two fundamental0-1 frequencies agree with the experimental absorption,as can be seen by projecting the 2D spectrum onto theW-axis. The cross-peaks are induced by pairwise vibra-tional couplings among local amide-I modes. The

Figure 11. 2D signals of a model system for the peptide bond (NMA)in the cross-peak region of amide-I, -II, -III, and -A modes. Top panel,left: Im[SkI

(�W1,t2 = 0 W3)] in the cross-peak region of amide-I, -II, and-III modes; bottom panel, left: Im[SkIII

(t1 = 0,W2,W3)] in the cross-peakregion of amide I, II, and III modes. Right panels show the samesignals in the cross-peak regions of the amide-A and amide-I, -II, and-III modes.

Figure 12. The amide-I–III kI cross-peak signals of NMA for different correla-tion coefficients. Left panels: Im[SkI

(W1,t2,W3)] signal for different h1,3 andh1,9 ; right panels: the actual simulated signals.


Chemie



diagonal and off-diagonal peaks, change upon 13C-labeling asshown in Figure 14 panels B and C. The spectra simulatedusing the HC model are shown in the middle row of Figure 14and the HM simulations in the bottom row. The main 2DIRcharacteristics of the UL, L2, and L7 are reasonablyreproduced by both models.

The effect of the multiple state non-adiabatic crossingbetween amide-I vibrational energy surfaces which isneglected in these simulations was investigated recently.[179]

5. Spectral Diffusion and Chemical Exchange:The Stochastic Liouville Equations

The Cumulant expressions used in Section 4 provide asimple compact description of bath fluctuations with Gaus-sian statistics coupled linearly to the frequencies. Moregeneral types of fluctuations require an expanded phase-space that includes relevant collective bath modes and tocompute the evolution of distributions in this extended space.This requirement can be accomplished using the stochasticLiouville equations (SLE) proposed by Kubo[180–183] to repre-sent the dynamics of the distribution of a quantum systemperturbed by a stochastic process described by a Markovianmaster equation. The SLE is widely used in the simulations ofelectron spin resonance (ESR),[184,185] NMR,[183] and IR[186,187]

line shapes.Below we apply this method to simulate 2DIR spectra and

the chemical exchange processes of a small peptide, triala-nine.[188] Trialanine has two amide bonds which contribute toits amide-I band. The Hamiltonian operator depends on thefrequencies wa and wb, anharmonicities Ka and Kb, and thecoupling constant J of the two local modes. The simulationspresented below include six vibrational energy levels: theground state (g), two single excited levels (e1 and e2) and threedoubly excited levels (f1, f2, and f3; see Figure 9 in Section 4).The time evolution of the density matrix describing the state

Figure 13. IR spectra of trpzip2 13C isotopomers; solid lines: UL;dashed lines: L2; dotted lines: L7. A) Experimental, B) simulated usingHC electrostatic potential model,[158] C) simulated using the HM multi-pole field model.[142]

Figure 14. A–C) Experimental kI + kII spectra of trpzip2 13C isotopom-ers. D-F) simulations using the HC Hamiltonian operator;[158] G–I) simulations results using the HM Hamiltonian operator.[142] Left: UL,middle: L2; right: L7. wt is W1 in the notation of this Review, while wt

is W3.




of the two mode system is described by the LiouvilleEquation (23).

L(t)1(t) =� i�h[H0(t),1(t)] represents the isolated system,

while Lint(t)1(t) =� i�h[Hint(t),1(t)] represents the coupling with

the radiation field.Analysis of the amide-I absorption band of triala-

nine[100,189–191] suggests that it primarily exists in the polygly-cine II (PII) structure (a conformation characterized byRamachandran angles of (y, f) = (�608, + 1408) and a righthand a-helix (aR) (y, f) = (�608, + 458).[192] We found 70%PII configuration and 30% aR in the joint distribution of theRamachandran angles derived by the MD trajectory. TheRamachandran-angle distribution functions for each config-uration were fitted to a Gaussian form. The two configura-tions are stable and only 38 transitions between themoccurred during the 10 ns simulation, suggesting an exchangetime of a few hundred picoseconds for the two species, whichis too slow to affect the line shapes. The response was thuscalculated as an inhomogeneous average over the two species.The non-adiabatic effect of the two-state curve crossings hasbeen also investigated.[179]

The frequency fluctuations of the two modes (dwa) and(dwb)are treated as independent stochastic variables. Theseare dominated by the interaction with the solvent watermolecules in the vicinity of each amide unit. The Brownianoscillator parameters (relaxation times ga

�1 = gb�1 = 220 fs

and magnitudes Da = Db = 16.1 cm�1 reproduce the experi-mental line shape for the isolated amide-I mode in NMA.[159]

The fundamental frequencies are given by wa = hwai+ dwa

and wb = hwbi+ dwb with average frequencies hwai= 1652 andhwbi= 1668 cm�1.[189,190] The difference stems from the chargeon the terminal amino group; the amide unit closest to theacid group has the lower frequency.

Ramachandran-angle fluctuations (df and dy) constituteanother set of relevant stochastic variables that primarilyaffect the intermode coupling J. J was expanded to quadraticorder to give Equation (24).

Cij were obtained by a fit to the TT map which connectsthe coupling constant and the Ramachandran angles.[146] C00

represents the coupling at the average Ramachandran angleswhich is the reference point for the Taylor expansion. Wefound C00 = 4 cm�1 in the PII configuration and 10.5 cm�1 foraR.

All four stochastic variables (dwa,dwb,df, and dy) aretreated as Brownian-oscillators, each characterized by twoparameters D(variance of fluctuations) and g (relaxationrate). The local anharmonicities defined as the differencesbetween the double of the fundamentals and the overtone

frequencies were fixed to 16 cm�1, their fluctuations wereneglected.[53, 98, 193] Transition dipole fluctuations of the localmodes were neglected as well and their magnitudes were setto unity.

The probability distributions P(Q,t) of our stochasticvariables Q1 = dwa, Q2, = dwb, Q3 = df, and Q4 = dy, ismodeled by the Markovian master Equation (25) whereG(Q) has the Smoluchowski (overdamped Brownian Oscil-lator) form [Eq. (26)].

The SLE is finally constructed by combining the Liouvilleequation for the exciton system [Eq. (23)] and the Markovianmaster equation [Eq. (25)] for the four collective Brownianoscillator coordinates [Eq. (27)].

The SLE may be solved using a matrix continued-fractionrepresentation of the Green functions,[188] in the frequencydomain. The 2DIR photon-echo signal SkI

(W1,t2,W3), wascomputed by transforming the frequency W2 back to the timedomain [Eq. (28)].

The Green function for the t2 interval may also becomputed in the time domain by a direct time integration ofthe SLE. Different levels of simulation of all parallel zzzzsignals were compared in Ref. [194]. The highest levelincludes fluctuations of all four collective bath coordinates.The Liouville operator is constructed in the local basis and thecoupling between the two local modes fluctuates with theRamachandran angles. The local-mode frequencies fluctuateas well. Satisfactory agreement with experiment is obtained asshown in Figure 15. Some differences arise since the aR

population is overestimated by the MD simulation. Stocket al.[195] demonstrated that different MD force fields predictvery different populations of the various conformations oftrialanine. However, the SLE need not necessarily rely onMD simulations and can use for example, parametersobtained from NMR spectroscopy.

In summary, four collective coordinates can account forthe effect of fluctuations on the two amide-I modes fortrialanine. Ramachandran angle fluctuations have significantsignatures on 2DIR line shapes in non-rigid peptides.


Chemie



The exchange between conformers in trialanine is slowand the signal is given by a sum of the contributions of thevarious conformers. Fast-exchange shows interesting signa-tures in 2D signals as demonstrated in hydrogen-bonding andisomerization dynamics.[196] These can be described byincluding a multistate jump model in the SLE. The Brow-nian-oscillator motion and the exchange process show differ-ent 2DIR signatures. In the following simulations we alloweda different width for the u and d peaks. The splitting 2D0 =

34 cm�1 (i.e. ca. 1.01 ps�1) and the exchange rates ku = 0.1 ps�1

kd = 0.125 ps�1 were taken from the cross-peaks growthreported in Ref. [136]. All three regimes were observedexperimentally in the formation and dissociation of phenol–benzene complexes in CCl4 solution.[136] In the intermediatetimescale regime (2 ps), memory of the Brownian oscillatorcoordinate is lost as evident by the circular line shape but thecross-peaks are weak. We thus assumed a L� 0.4 ps�1

relaxation rate. L,W1, and W3 can be estimated from theabsorption linewidth using the Pade approximate of a two-level system.[66] Using W1 = 0.33 ps�1, W3 =�0.07 ps�1 simu-lations reproduced the experimental absorption spectra. The2DIR–photon-echo signals shown in Figure 16 recover allexperimental features; all three regimes are clearly seena) rephasing elliptic shapes, b) the relaxed Brownian oscil-lator with circular shape, and c) chemical-exchange cross-peaks as found experimentally (the lower frequency peak isweaker but broader[136]).

The SLE can be used to describe many types of fluctua-tions of all the elements of the Hamiltonian operator. Theonly requirement is that they can be represented by a few(discrete or continuous) collective coordinates that satisfy aMarkovian equation of motion. These equations account forthe effect of the fluctuations of collective bath coordinates onthe nonlinear IR spectra by describing the evolution in thejoint system-and-bath space.

6. The OH Stretch Band of Liquid Water

Liquid water has many unique properties stemming fromits unusual capacity to form multiple hydrogen bonds, makingit the most important solvent in biology. These bonds andtheir fluctuations has been extensively studied.[92, 121, 175,197–206]

The vibrational OH stretch band is complicated byresonant exciton transfer to neighboring mole-cules.[175,204, 207–210] The spectrum of HOD in D2O has receivedconsiderable attention since it is a simpler model systemwhere such transfer is not possible (OH frequency is3400 cm�1, OD frequency is 2500 cm�1).[211] The absorptionbandwidth of the OH stretch of the HOD/D2O

[120,212, 213] is255 cm�1 (full width at half maximum height (FWHM))[212]

and shows a 307 cm�1 [212] solvent red shift from the gas phasefrequency 3707.47 cm�1.[214] A 70 cm�1 vibrational Stokes shiftin IR fluorescence was reported by Woutersen andBakker[215–219] Vibrational relaxation and hydrogen-bonddynamics were also probed by spectral hole burning, two-pulse photon-echo experiments, and photon-echo peakshift.[120, 213, 220,221] An observed oscillation was attributed to acoherent hydrogen-bond motion, as verified by simula-tions.[120, 222] Similar photon-echo experiments and simulationswere carried out on a complementary system (OD stretch ofHOD in H2O).[122, 223] It was recently proposed that the fifth-order nonlinear IR experiment (3D-IR)[224] can monitor thethree-point frequency fluctuation correlation function,revealing the relation between the spectroscopic coordinatesand dynamic coordinates of hydrogen-bond rearrange-ments.[225]

The electrostatic ab initio map approach described inSection 3 was employed to simulate the O�H stretch funda-mental and its overtone.[226] The anharmonic vibrationalpotential of HOD expanded to the 6th order in the threenormal coordinates (H-O-D bending, O-D stretch and O-Hstretch) in the multipole electric field were calculated at theMP2/6-31 + G(d,p) level. Simulated CGF solvent-inducedpeak shift and bandwidth (Figure 17; 287 cm�1 and 309 cm�1)are in good agreement with experiment (306 cm�1 and250 cm�1; Figure 17)

A collective electrostatic coordinate (CEC) W wasintroduced for the O-H stretch; a linear combination of themultipole electric-field coefficients which is defined as a

Figure 15. Top: The experimental kI photon-echo spectrum of trialani-ne[194] (left) and the simulated spectrum (right) for parallel polarizedpulses. Bottom: Same comparison but for perpendicular polarizedpulses. The spectra are normalized to the most intense peak.

Figure 16. Simulated 2DIR signals SA = kI + kII for exchangeW1 = 0.5 fs�1 L = 0.4 ps�1; W2 = 0.33 ps�1, W3 =�0.07 ps�1

kd = 0.125 ps�1, ku = 0.1 ps�1; D0 =�2.0 ps�1; D3 = D1 = 0. Time delays:a) t2 = 0; b) t2 = 2 ps; c) t2 = 10 ps. These spectra closely resemble theexperimental results of Ref. [136].




linear part of the electrostatic frequency map [Eq. (6) inSection 3.1] in C around the average hCi [Eq. (29)].

We use the coordinate system shown in Figure 17. Thefrequency fluctuations are well described by a quadraticpolynomial in W [Eq. (30)].

The scatter plot of the frequencies calculated withselected electrostatic components versus the full componentcalculation given in Figure 18 shows that three (Ez, Ezz, and

Exx) components dominate the overall frequency shift fromthe gas phase. The frequencies calculated with only Ez

(Figure 18 left panel) are systematically too high, indicatingthe significant contribution of Ezz and Exx to the O-H stretchfrequency. Exx is dominated by the hydrogen bonding ofoxygen of HOD to the deuterium of D2O solvent. The partialcharge of deuterium in D2O creates the diagonal negativegradient of the out-of-plane electric field, and the simulatedensemble average values of hExxi (�0.0094) verifies this point.

The CEC correlation function shows biexponential decay.The CEC was therefore decomposed into a sum of a fast (W1)Brownian oscillator coordinate representing hydrogen-bond-ing fluctuations and a slow (W2) coordinate representing thesolvent fluctuations outside of the first solvation shell (W =

W1 + W2). The relaxation times for W1 and W2 are t1 = 34.4 fsand t2 = 0.501 ps. Two stochastic models, the collectiveelectric coordinate (CEC) and four state jump (FSJ) wereemployed for simulating the effects of hydrogen-bondingfluctuations W1 on the line shapes.[226, 227] The CEC modelassumes two CEC (W1 and W2) which describe fast and slowfluctuations assuming the continuous Gaussian processes. TheFSJ model uses a master equation to describe the jumpsbetween four hydrogen-bonding configurations in addition tothe slow CEC fluctuation (W2). Twelve hydrogen-bondingconfigurations were obtained by employing the geometrichydrogen-bonding criteria.[203, 228,229] These were clustered intofour groups, configuration I (one hydrogen bond to eachhydrogen and two to oxygen), II (one hydrogen bond to eachhydrogen and less than two to oxygen), III (no hydrogen bondto hydrogen, but two hydrogen bonds to oxygen), and IV (nohydrogen bonds to hydrogen and less than two hydrogenbonds to oxygen).

While the CGF band shape is symmetric, the CEC modelpredicts an asymmetric band with a long red tail, consistentwith experiment (Figure 17).[120] The anti-diagonal linewidthsof photon-echo signal in CGF and CEC are about the same atlow and the high frequency (Figure 19). However the fourstate jump linewidth is 23 cm�1 larger for higher frequencies,despite the fact that the frequency distribution is broader forthe low-frequency configuration I. The blue section in theexperiment as marked in Figure 19 is 19 cm�1 broader thanthe red section. We define the symmetry parameter h in termsof the FWHM line widths of the red (GR) and the blue (GB)anti-diagonal slices (Figure 19): h = (GB�GR)/(GB + GR). TheFSJ asymmetry parameter h (0.125; Figure 19 top row, middlepanel) is in better agreement with experiment (0.0848) [230]

than CEC (0.0138; Figure 19 top row, left). The CGFsimulation gives a symmetric band shape along the diagonalblack line (Figure 19, bottom) and misses the observedasymmetry.

The experimental h suggests that the shorter lifetime ofthe high-frequency hydrogen-bond species gives rise to aconsiderable line broadening. Hydrogen-bond kinetics aremuch faster than the slow dynamics responsible for thefrequency distribution of the individual species. The triangu-lar shape of the diagonal photon-echo peak can be attributedto fast femtosecond hydrogen-bonding kinetics. In the FSJmodel, breaking a hydrogen bond on oxygen affects both thefundamental OH frequency and the anharmonicity more thanbreaking the hydrogen bond on the hydrogen atom. Hydro-gen bonding to deuterium causes a blue shift.

We define the anharmonic shift as the frequency differ-ence along the w3 axis between the peak positions of thestimulated emission and the ground state bleach peak.Anharmonic fluctuations add 10 cm�1 to the anharmonicshift. Hydrogen bonding to the H atom of HOD lowers theOH stretch vibrational potential of HOD more at longer O�H bond lengths, making the anharmonicity larger. Water in

Figure 17. Simulated linear IR O�H stretch line shape calculated withthe CGF (thin solid line) and the SLE (thin dashed line) as well as theexperimental data (thick solid line).[120] The black vertical arrow repre-sents the gas-phase frequency.[214]

Figure 18. Scatter plots of the frequencies calculated with variouselectrostatic components versus the full calculation. Crosses representthe gas-phase frequency,[214] and diagonal lines represent the perfectagreement. Left: Ez ; right: Ey, Ez, Eyy, and Exx. All axes are in cm�1.


Chemie



confined environments (membranes, interfaces, reversemicelles) can be effectively studied using 2DIR.[231–233] Even-order signals R(2) and R(4) vanish in isotropic systems and thusprovide very sensitive probes for interfaces.[234–236] Sumfrequency generation is a 1D technique. Multidimensionalextensions are on the horizon.[237]

Simulations of 2DIR spectra of neat liquid water (H2O)must account for highly disordered coupled-resonant O-Hstretch vibrations. In addition to the modulations of thetransition frequencies, which also exist in the HOD/D2Osystem, dipole moments, and anharmonicities fluctuations ofthe intermolecular coupling in the extended hydrogen-bondnetwork now become relevant. The first photon-echo stud-ies[123,175] on neat H2O have revealed significantly fasterstructural dynamics than HOD in D2O. This result wasattributed to a stronger coupling to librational motions, withpossible contributions from resonant energy transfer anddelocalization of the vibrational excitations.

Simulations of the 2DIR photon-echo and pump–proberesponse of the O-H stretch vibrations of liquid water[238] wereperformed by a direct numerical integration of the Schr�-dinger equation, including both symmetric and antisymmetricstretches, intermolecular couplings, as well as fluctuations andanharmonicities of transition frequencies and dipolemoments. This simulation allows for multiple-state non-adiabatic crossings between vibrational energy surfaces onany time scale.

The dielectric constant was used to scale the resonantdipole–dipole coupling and reproduce the observed polar-ization anisotropy decay (80 fs). The next-neighbor couplingstrength (12 cm�1) gives the best agreement. A lifetime of200 fs is assumed. Figure 20 shows that the simulated peakshapes, amplitudes, and dynamics are in close agreement withexperiment. The negative and positive peaks correspond tothe fundamental transition and the excited-state absorption,

Figure 19. Comparison of the 2DIR photon-echo spectra, calculated using the two SLE and two CGF models, with the experimental data. The fullblack line is the diagonal, the dashed line is displaced 100 cm�1 above the diagonal. The red and blue lines show where the anti-diagonal slicesare taken on the red and blue side, respectively, to calculate the asymmetry parameter h. CEC(i): SLE simulation using the CEC; FSJ: SLEsimulation using the FSJ; CGF(i): CGF simulation; CGF(ii): CGF simulation with an infinite negative anharmonicity.

Figure 20. 2DIR photon-echo spectra (kI =�k1 + k2 + k3) of the O�Hstretch vibration in H2O for population times t2 = 0, 50, 200 fs. Top:experimental data,[123] bottom: simulations using a direct numericalpropagation. Each spectrum is normalized to its maximum. (adaptedfrom Ref. [238]).




respectively. In both experiment and simulation, the funda-mental peak is stretched along the diagonal, indicating someinitial inhomogeneity at t2 = 0 fs. As t2 is increased, the peakorientation becomes more vertical. The bending of thefundamental peak and the nodal lines between the twopeaks indicates faster fluctuations and loss of inhomogeneityon the red side of the spectrum. Initial correlations on the redside of the spectrum decay in 100 fs, but persist beyond 200 fsin the blue side.

The large number of acceptor modes, as well as anharmo-nicities and fluctuations in water open up many intermolec-ular transfer pathways, which lead to a full decay of thepolarization anisotropy on observed time scales (80 fs) eventhough the average coupling (12 cm�1) is weak. The effect ofresonant energy transfer on the 2DIR photon-echo spectra isfound to be rather small for short t2 time (< 200 fs). Most ofthe fast dynamics in the 2DIR photon-echo spectrum arecaused by the local O-H stretch frequency fluctuations owingto the sensitivity of the local anharmonic potential to thefluctuating hydrogen-bonding environment. The O-H stretchvibration is an excellent probe of the hydrogen-bond networkin H2O. 2DIR of other liquids, such as formamide, may besimulated using the same approach.[239–241]

The study of water dynamics in confined biological,[249,250]

chemical,[251] and geological[252] environments is of consider-able theoretical and experimental interest.[231–233, 242–248]

Considerable 2DIR activities had focused on reversemicelles, in particular aerosol OT (AOT). Studies of the OHstretch absorption of diluted HOD in a droplet of D2O orH2O

[242–245] have shown that the dynamics of the confinedwater is slower than in the bulk. Using stimulated vibrationalecho and spectrally resolved vibrational echo peak shift,Fayer and co-workers have shown that the fastest dynamicsresulting from hydrogen-bond-length fluctuations (50 fs) inconfined water is similar to the bulk, but the timescale of theslower global structural evolution (> 1 ps) could increase byan order of magnitude in strongly confined systems.[244, 245] Thedynamics of neat water in reverse micelles has been reportedas well.[231,233, 246] IR pump–probe and vibrational-echo spec-troscopy support the existence of two independent relaxingwater sub-ensembles.[231,233] The dynamics of the core of thedroplet is similar to the bulk, but the shell is slower. A 0.4 nmshell thickness has been measured.[231] The confinement ofwater in phospholipids membranes has been studied aswell.[232, 247,248] In contrast to reverse micelles where theconfinement induces a core–shell separation, water dynamicsin membranes is dominated by strong hydrogen bonds withthe phospholipid polar groups.

7. Application to Phospholipids: QuasiparticleRepresentation of 2DIR Signals

As a constituent organelle in the cell,[253] the membranesets the information and energy gradients and controls theirflow, which is essential for life. The common structuralmoieties in the polar surface of cellular membranes, carbonyl,phosphate, and chlorine mediate molecular recognition andsignal transduction.[253, 254] Owing to experimental limitations,

our knowledge of their arrangement and dynamics is not verydetailed. In a lipid bilayer, lateral irregularities smear theneutron scattering diffraction pattern, and NMR resonancesare broad because of the restricted motions which result inincomplete motional narrowing, as found in solid-state NMRspectroscopy. IR spectra of the carbonyl moieties in phos-pholipid membranes has attracted considerable atten-tion.[255–259] The absorption-band shows a clear inhomogene-ous character and can be described as a superposition ofseveral sub-states.[255, 256] Carbonyl stretching line-shapes inphospholipids could yield direct information about moleculararchitecture and fluctuations in the membrane inter-face.[260, 261] There are many sources for the high spectralinhomogeneity differences found in the local environment ofthe sn-1 and sn-2 carbonyl moieties, these stem from thepacking arrangements,[256,258] local chain conforma-tions,[258, 259,262–264] the relative positions of the two carbonylswith respect to the interface,[132,259, 265] and the degree ofhydration.[266, 267] In an elegant study, Blume et al.[266] ruled outall scenarios involving local structural differences excepthydrogen bonding. Another study[268] similarly eliminated thevariance in hydration as a possible source of inhomogeneity.

Figure 21 shows a bilayer of the phospholipid dimyris-toylphosphatidylcholine (DMPC) [107] The 2DIR of the C=Ostretch band of this bilayer is shown in Figure 22. The sn-1 andsn-2 carbonyl degeneracy is lifted by a 13C labeled carbon inthe sn-2 chain, giving two broad vibrational bands at 1740 and1697 cm�1.

The SOS method described in Section 3 requires thediagonalization of the two-exciton block of the Hamiltonianmatrix. The N4 scaling of time makes computation prohib-itively demanding for large N. An alternative, quasiparticlescattering, approach greatly reduces the computationaldemand. This method assumes a molecular Hamiltonianoperator that conserves the number of excitations, and adipole moment that can only create or annihilate oneexcitation. The optical transitions are viewed as quasiparticles (“excitons”), and the nonlinearity now originatesfrom their collisions.

The amide-I Hamiltonian [Eq. (3)] can be approximatelyrecast in terms of Bosonic creation and annihilation operators

Figure 21. Chemical structure of the DMPC phospholipid and a snap-shot of the DMPC bilayer taken from the MD simulation (for claritythe water molecules are not shown). Orange P, blue N, and red O. Thehydrophobic tails are shown with sticks.


Chemie



[Eq. (31)], where B† and B are creation and annihilationBoson operators, respectively, satisfying [Bm

†,Bn] = dnm. The

first two terms describe the free excitons where wm is the localamide-I frequency and the Jmn represents the inter-sitecoupling which induces exciton hopping. Dm and Kmn repre-sent the intra- and inter-site anharmonicity, respectively.

The quasiparticle picture appears naturally by solvingequations of motion, the nonlinear exciton equations(NEE).[62, 64] A key ingredient is the Green function, g,which describes the time evolution of two excitons andsatisfies the Bethe–Salpeter Equation (32).

g(0) represents the dynamics of two non-interacting

excitons. G is the two-exciton scattering matrix. Its matrixelement Ge4e3,e2e1

W[61] represents a process where two incomingexcitons e1 and e2 are scattered to produce e3 and e4.

Formally the computational effort of both the quasipar-ticle expression and the SOS scale as N4 with system size.However, a much more favorable scaling is obtained in

practice thanks to the localized nature of excitons and theirinteractions (anharmonicities).[55] To see how this works, theoverlap factor of two excitons needs to be considered[Eq. (33)].

This parameter characterizes the two-exciton configura-tion in real space: for e = e’ we have hð1Þee0 � 1. For uncoupledvibrations Jmn = 0 and hð1Þee0 = dee’ indicating that the excitons donot interact. Since exciton interactions are short-range (theanharmonicity is local) we can estimate the probability of thescattering event by assuming that exciton pairs (e1e2) can onlyscatter when their overlap is larger than a certain cut-offhc :h(1)

e1e2>hc. The same criterion may also be used for the

pairs of outgoing states (e4e3). By applying this cut-off whichrestricts the distance between two initial and between twofinal excitons in the scattering matrix the number of relevantscattering matrix, elements should scale as N2Nc

2 rather thanN4, where Nc is a finite correlation length; the scatteringmatrix is sparse.

A second helpful constraint is provided by the exciton–exciton scattering radius which determines how far twoexcitons can move during their interaction, and sets boundson the distance between the initial and final pairs of excitons.We introduce this cut-off by defining a second overlapparameter [Eq. (34)].

h(2) is the amplitude of a path going from e to e’ through allpossible intermediate states e1. A cut-off of hð2Þee0 may be used toselect the dominant e3e2 pairs in the scattering matrix Ge4e3,e2e1

.Using both cut-off parameters hð1Þc and hð2Þc , we can retain

only those scattering matrix elements which satisfy hð1Þe2e1>hð1Þc ,

hð1Þe4e3> hð1Þc , hð1Þe3e2

> hð2Þc , hð1Þe3e1> hð2Þc , hð1Þe4e2

> hð2Þc , and hð1Þe4e1> hð2Þc . The

scaling of the NEE effort with system size thus reduces to N.The reduction in computational effort, which becomes morepronounced as the system size is increased, stem from twofactors: 1) The relevant exciton states may be identifiedbefore calculating the scattering matrix. Their number istypically much smaller than N4. The scattering matrix shouldbe calculated only for the selected set of scattering config-urations. 2) The required numerical effort for computing thesignal using multiple summations is reduced considerably bythe sparse nature the scattering matrix.

Figure 22 shows the experimental and simulated pump–probe spectra of carbonyl moieties in a phospholipid bilayerfor parallel and perpendicular polarization configuration ofthe pump and the probe pulses. The local amide-I frequenciesare 1708 cm�1 (13C labeled) 1755 cm�1 (unlabeled) correctedby a Stark effect frequency shift: Dw = kEproj, where Eproj isthe projection of the electric field along the C=O bond. Off-diagonal elements were obtained by using the transitiondipole coupling model.[149] The experiment uses a spectrally

Figure 22. Left column: experimental pump–probe spectra of thecarbonyl groups in phospholipid membrane fragments. From the top:absorption, and pump–probe spectra recorded under parallel andperpendicular polarization conditions of the pump and probe pulses.Right Column: The corresponding simulated spectra. Blue andmagenta brackets show the inter- and the intra-band cross-peakregions, respectively.[107]




narrow (16 cm�1) pump and a short (100 fs) impulsive probe.The signal field is spectrally dispersed. The experimentalbandwidths resulting from the fluctuating electrostatic envi-ronment as well as their (diagonally-elongated) shape char-acteristic to inhomogeneous broadening are reproduced bythe simulations. The two strong diagonal resonances corre-spond to absorption by the two carbonyl groups.

The cross-peak regions in the 2D signal are weak. The twohorizontal sections of the calculated and experimental signalsare compared in Figure 23. The intensities and line shapes of

both intra- and inter-band cross-peaks are fairly well repro-duced. Each resonance has a negative (blue) contributionowing to GSB and ESE, and a positive (red) ESA contribu-tion (see Figure 10). The red shift of the ESA band reflects theanharmonicity of the carbonyl stretching mode. The cross-peaks are more pronounced when the pump and the probehave perpendicular polarization (Figure 22, ? ). Figure 23Bdepicts horizontal (k ) and perpendicular ? sections ofFigure 22 at the pump frequencies 1675 and 1752 cm�1

(marked by arrows). The cross-peaks provide a directmeasure of vibrational coupling between carbonyl moieties.Structural information, such as the distribution of anglesbetween intramolecular carbonyl pairing, may be obtainedfrom quantitatively comparing the simulated and experimen-tal results. The pairing geometry is expressed in terms of theangle between the transition dipole moments and theirseparation. The vibrational frequencies of the two coupledcarbonyl groups wn and wm are obtained by diagonalizing theexciton Hamiltonian operator [Eg. (35)].

These depend on the coupling parameter Jmn and on thedifference of the diagonal frequencies, w0

n�w0m. We further

define the pair coupling parameters b’mn [Eq. (36)] and theweighted radial angular pair distribution function [Eg. (37)],where the m and n sums run over the 12CO and 13CO carbonylgroups, respectively.

Figure 24 shows h(R,q) calculated by considering all12CO:C13CO pairs (A), only the intermolecular pairs (B),and only the intramolecular pairs (C). Panel A in Figure 24shows that h(R,q) vanishes for distances of > 6.5 � implying

that the cross-peaks are dominated by neighboring carbonylgroups. The distribution function h(R,q) (Figure 24A), con-sists of several structural families whose intermolecular orintramolecular origin can be easily traced by comparison withFigure 24 B and C. The intermolecular h(R,q) does not showrandom orientations even when it is broader than its intra-molecular counterpart. The sharp peak at q = 408 and R = 5 �in Figure 24 C is in agreement with the angle between thetransition dipole moments obtained from the experimentalanisotropy, suggesting that it is mainly due to intramolecularpairs. We note that for this angle, intermolecular carbonylpairs also contribute significantly (up to 26� 5 % to the totalh(R,q) function (see Figure 24).

These simulations reveal the important role of electro-static interactions at the polar interface. Both the transitiondipole moment coupling and the electric-field fluctuationsaffect the absorption band line shape. The two contributions,which are convoluted and indistinguishable by the linearresponse, can be clearly separated in the diagonal and in theoff-diagonal parts of the 2D correlation plots. The cross-peakintensity provides a direct measure of the contribution ofcoupling to the overall line shape. The diagonal elongationresults from both the frequency dispersion of the excitonicstates and the local electric field fluctuations. The 2D line

Figure 23. A) experimental absorption (thick line) and calculated linearoptical absorption (thin line) of DMPC in water. B) Experimental (opencircles) and calculated (solid lines) hole-burning spectra under pumpexcitation at 1675 and 1752 cm�1 (see arrows). Black (parallel) andred (perpendicular) colors indicate the polarization conditions. Theperpendicular spectra are magnified by a factor three. Blue andmagenta brackets mark the inter- and intra-band cross-peak regions,respectively (see also Figure 22).

Figure 24. b’-weighted radial-angular distribution functions [Eq. (37)]of the simulated DMPC bilayer, calculated considering A) all 12CO–13CO pairs, B) 12CO–13CO intermolecular pairs, and C) 12CO–13COintramolecular pairs. The red dotted lines indicate the angular valuesobtained from experimental spectral anisotropy. The chromatic barshows the range of the statistical distribution according to Equa-tion (37).


Chemie



shapes provide a unique window into the vibrational excita-tions. The increased degree of localization of the excitonicstates in the absorption tails reflects local structural proper-ties of the nearest chromophores.

2DIR combined with quasiparticle simulations provide apromising structural tool for studying composite phospholipidbilayers, host–guest lipid–protein complexes, lipid systems ofreduced dimensionality, and polymers.

8. Double-Quantum-Coherence Spectroscopy

Elaborate pulse sequences are routinely designed in NMRspectroscopy to extract desired information. Similarly inter-ferences between quantum pathways underlying multidimen-sional signals may be manipulated to design new 2DIRtechniques. Herein we demonstrate a signal designed tovanish for non-interacting excitons thereby providing anexcellent probe for such interactions.

The applications presented so far focused on the kI =

�k1 + k2 + k3 and kII = k1�k2 + k3 signals. The kIII = k1 +

k2�k3 signal carries different types of information. It isgiven by the two quantum pathways ESA1 and ESA2(Figure 25), analogous to the double quantum coherence

technique in NMR spectroscopy.[269] In both diagrams thesystem is in a coherent superposition of the doubly excitedstate f and the ground state g during t2. This time-interval thusprovides a clean view of two-exciton states. We shall consider(W2,W3) 2D spectra obtained by varying the t2 and t3 delays.(W1,W2) signals are also possible.

As W2 is scanned, the signal shows resonances corre-sponding to the different doubly excited states f. However, theprojection along the other axis (W3) is different in the twodiagrams. In ESA2 the system is in a coherence between e’and g during t3. As W3 is scanned, it reveals single excitonresonances when W3 = we’g. For ESA1 the system is in acoherence between f and e’ during t3. This situation gives riseto many new resonances at W3 = wfe corresponding to all thepossible transitions between doubly and singly excited states.The remarkable point is that for non-interacting excitons thestate f is simply given by a direct product of the single pairstates e and e’, the double-excitation energy is the sum of the

single-excitation energies, and the two diagrams exactlycancel. The resonance pattern of these 2D correlation plotsprovides a characteristic fingerprint for the correlated natureof two excitons.

The enhanced resolution of kIII signals stems from theabsence of diagonal peaks which dominate the kI spectra andcover the off-diagonal (cross) peaks, and from the doubledfrequency bandwidth of two quantum coherences.

We demonstrate that (W2,W3) correlation plots of kIII forthe 74-residue TB6 protein domain (Figure 26)[270] are more

sensitive to the couplings between vibrational modes, com-pared with (W2,W3) correlations in kI. We have used sensitivityanalysis to assign various regions in congested spectra ofglobular proteins to specific secondary structures and toseparate the overlapping regions. We add a small shift hn tothe energies en

m of all modes belonging to the v’th secondarystructure motif (hn should be much smaller than all Jmn). Thedifference of the perturbed and the unperturbed spectrumreveals its sensitivity to this perturbation, and its spectralregion can then be assigned to the structure of type n.Figure 26 gives the simulated kI (SkI

)and kIII (SkIII) signal and

shows the dissected signal related to helix and hairpinsegments and their couplings.

9. Chirality Effects: Enhancing the Resolution

Pulse polarizations provide a whole host of convenientcontrol-parameters that may be easily varied to manipulatethe 2DIR signals. We label a coherent heterodyne third-order

Figure 25. Double-sided Feynman diagrams representing the Liouvillespace pathways contributing to the signal in the rotating-waveapproximation. The first excited-state absorption (ESA1) diagramcorresponds to R7 and the second excited-state absorption (ESA2)diagram to R’4.

Figure 26. Top row: Simulated signal and sensitivity analysis for the kI

signal of TB6 protein domain. SkIis the signal. ab(kI) gives the regions

related to a helix (red contour) and b sheet (black contour). Jab(kI)gives the region related to the coupling between a helix (red contour)and b sheet. Bottom row: same quantities for the kIII signal.




signal nsn3n2n1, where the three incoming pulses are polarizedalong the n1,n2,n3 directions and the signal is polarized alongn4, as (Figure 27).

Molecules are typically smaller than the optical wave-length and their response may be adequately described byassuming that the field is uniform across the molecule; this isknown as the dipole (or long wavelength) approximation. Ouranalysis so far was restricted to this limit. We further assumedthat all pulses are polarized in parallel and did not need tospecify the pulse polarizations. The nonlinear responsegenerally depends on the orientationally averaged productof four dipoles hmns

mmn3n mn2

k mn1l i. In isotropic samples there are

only three independent polarization configurations: xxyy,x-yxy, and xyyx. All other configurations can be expressed bytheir linear combinations.

The variation of the phase of the optical field at differentpoints within the molecule may result in new contributions tothe signal. These are caused by interferences among signalsgenerated at different parts of the molecules and are typically1000-times weaker than the leading (dipole) contributions(this is the ratio of chromophore size to the optical wave-length). However, by choosing polarization configurations forwhich the dipole term vanishes (e.g. xxxy), the non-dipolesignals are background-free and may be readily detected.These signals change their signs upon mirror reflection; hencethey vanish in racemates and in nonchiral molecules and onlyexist in chiral systems.

Circular dichroism (CD), the difference in the absorptionof left- and right-handed circularly polarized light,[271–273] is thesimplest chiral signal. This linear 1D technique is routinelyapplied for probing the folding states and conformations ofproteins. CD spectra have positive and negative componentsand the contributions of different chromophores interfere(the absorption spectrum in contrast is positive and additiveand contains no interference). This property is the reason forthe extra sensitivity to structure, allowing the CD technique todistinguish between various secondary structures of proteins.Similarly the structural sensitivity of 2D techniques can begreatly enhanced by a judicious choice of chiral polarizationconfigurations. Chirality-induced (CI) 2D techniques are

extensions of CD to nonlinear spectroscopy.[274, 275] Chiralitycan also be measured by the Raman optical activity technique(ROA),[276] which measures the difference in the Ramanintensities induced by right and left circularly polarizedincident light. Vibrational CD (VCD) band shapes arecharacteristic to secondary structures of polypeptides. Pro-tonated a-helical structure give bisignate amide-I and amide-A bands,[277, 278] and a monosignate amide-II band.[279] Theamide-I has three peaks for right-handed helices upondeuteration.[279] a-Helix and antiparallel b-sheets are distin-guishable by the amide-I frequency shift and the band shapeschange from the bisignate form with a small peak splitting totwo well-separated negative peaks.[280] The amide-I bandshape of random-coil structures is also bisignate, but its sign isreversed compared to the a-helix.[280,281] The 310 helix has ahigher frequency amide-II VCD band and a lower frequencyamide-II IR band compared to the a-helix. This difference isdue to the difference in hydrogen-bond patterns (4!1 vs 5!1).[282]

The response function for a chirality-induced kI techniquedepends on the averaged product [Eq. (38)][64] where jn

e(k) isexciton transition dipole in k space [Eq. (39)].

Molecular chirality is recast to the three-dimensionaldistribution of local transitions in real space. For simplicity,we neglect the local chirality of each peptide unit (because oftheir magnetic dipole and electric quadrupole) and onlyinclude the global (structural) chirality. Signals sensitive tochirality depend explicitly on the positions of the various localtransitions. We define the transition dipole vector for thezero-momentum exciton state as in Equation (40) and thefirst-order contribution in k as in Equation (41).

In these equations, rm is the coordinate for the m’thtransition, mm is the transition dipole, and ye,m is the excitonwavefunction. Equation (40) is independent of rm and insen-sitive to chirality. Equation (41) goes beyond the dipoleapproximation. For components such as xxxx with an evennumber of repeating indices, the first term is finite and willdominate the signal, making it insensitive to chirality. Forcomponents with an odd number of repeating indices such asxxxy, the first term vanishes and the signal is dominated by theother chiral-sensitive terms. The chirality-induced signalsdepend on products of the form hrn5

mnmn4mmn3

n mn2k mn1

l i. Nonchiraltechniques depend only implicitly on the structure through itseffect on the frequencies and transition dipoles which

Figure 27. Pulse configuration for femtosecond coherent IR correlationspectroscopy. Three laser pulses (light blue) interact with the sample.The fourth pulse (red) is used to detect its nonlinear response. Thecontrol parameters are the time intervals between pulses t1,t2,t3. All thepulses propagate along the z direction (collinear). The nonchiral signalxxxx is generated when all the pulses are polarized along x (blue andred). The chirality induced xxxy signal is obtained by switching the firstpulse polarization direction to y (green).


Chemie



influence peak positions and intensities. The explicit coor-dinate dependence of the chiral response amplifies the cross-peaks and is the reason why these techniques are moresensitive to fine details of the structure.

In isotropic samples there are three independent chirality-induced polarization configurations for collinear pulses andsix additional non-collinear terms.[61] The signals furtherdepend on the magnitudes and directions of pulse wave-vectors. Figure 28 shows the simulated IR chiral response ofthe amide-I vibrations where all beams propagate along z.The electronic CD spectra given for comparison were

simulated using Woody�s standard exciton model whichincludes the electric and magnetic moments of the chromo-phores.[271] It thus depends on both local and global chirality.Our simulations show how the CI techniques providecomplementary information to CD and NMR spectroscopyfor a 15-residue hairpin Trpzip4 (Figure 28A),[283] one of the“Tryptophin Zipper” hairpins. Its robust structure makesTrpzip4 an excellent model for the characterization of thevibrational states of peptides in aqueous solution, for theinvestigation of the relations of the vibrational spectra withpeptide conformations, and for the evaluation of the distri-butions of structures.[283] The amide-I absorption band (Fig-ure 28B) consists of three overlapping features; the 1635 cm�1

peak and the 1675 cm�1 shoulder are related to the b struc-ture,[283] while the 1655 cm�1 shoulder is related to the turnand coil structures at the two ends. The diagonal peaks of 2Dxxxx signals (Figure 28 C) resemble the absorption. NMR

spectra are routinely used for imposing constraints on peptidestructure; a distance geometry optimization is then applied toobtain an ensemble of possible conformers consistent with theNMR data.[8] We focus on the first two conformers out of the20-reported NMR-spectroscopy determined Trpzip4 struc-tures, which have the lowest energy and are thus the bestapproximation of the structure. The RMSD between thesetwo structures is 1.517 � The calculated electronic (Fig-ure 28D and G) and vibrational CD spectra (Figure 28Eand H) of these conformers are similar. However, the 2Dchirality-induced spectra (Figure 28 F and I) are very differ-ent. Conformer I has a strong (1635 cm�1, 1655 cm�1) cross-peak while II has cross-peaks (1655 cm�1, 1675 cm�1).

These examples demonstrate how chirality-induced 2Dsignals can help determine correlations between differentparts of a protein by enhancing certain cross-peaks therebyallowing their assignments to structural features. The cross-peaks are very sensitive to secondary structure variations, andthe chiral configuration of different chromophores can bedetermined from the signs of the corresponding cross-peaks(positive vs. negative cross-peaks between two transitionscorrespond to different sense of screw configuration of thecorresponding transition dipoles). Coherent 2D techniquesenhanced by the spatial sensitivity of chirality-induced polar-ization configurations offer a powerful tool for tracking earlyprotein folding events and pinpointing the average structureand its fluctuations along the folding pathways with femto-second resolution.

Chirality-induced 2D signals are weaker than their non-chiral counterparts, and have not been observed experimen-tally to date. Nevertheless since they are background-freethey may detected using state-of-the-art IR technology. Thecollinear pulse configuration presented herein is the simplestsuch technique. The wavevector selectivity of various techni-ques is missed in this case, but it can be recovered using phase-cycling techniques as done in NMR spectrosco-py.[23–25, 48–55,63, 269, 284] Combinations of several carefullyarranged non-collinear experiments may lead to the cancel-lation of nonchiral terms, so that only the chirality-inducedterms survive. The pulse configuration may be tailored forprobing specific tensor components. For instance, the col-linear xxxy signal can be measured in non-collinear geometrywhereby all the laser beams are arranged in one (yz) plane,the first y-polarized beam propagates along z and the other x-polarized beams can have wavevector component along y. Allnonchiral contributions vanish for this configuration and onlyxxxy survives.

10. The Structure of Amyloid Fibrils

The accumulation of amyloid deposits,[285] whose domi-nant component is a 39–43 residue Ab peptide,[286] has beenidentified as a major feature of the pathogenesis of Alzheim-er�s disease (AD).[287] Despite their identical 1–39 sequence,the various Ab peptides have significantly different biochem-ical properties: The 42-residue derivative Ab42 depositsmuch faster than others and the fibrils formed are morestable.[288] Ab 42 is also slightly more hydrophobic, compared

Figure 28. Using chirality-induced 2D correlation spectroscopy to dis-criminate between the hairpin structures that are indistinguishable byNMR spectroscopy. A) Fifteen-residue hairpin-peptide Trpzip4. B) Sim-ulated (red) and experimental (green)[17] absorption of the amide-Ivibrational band. C) Simulated xxxx 2D signals for the amide-I band.Middle and bottom rows: Comparison of the simulated spectra fortwo configurations drawn from the NMR-determined hairpin-structureensembles. D and G) Electronic CD spectra of the amide band, E andH) vibrational CD of the amide-I band, and F and I) xxxy chiralityinduced 2D signals for the amide-I band. The CD signals are similarfor the two configurations shown. Major differences of the 2D signalsin the cross-peak region indicate specific couplings among vibrationalmodes.




with shorter analogues, such as Ab40, because of the twoadditional more-hydrophobic residues at the end of thepeptide strand.[289] More importantly, the protease resistanceof Ab 42 is drastically different from its analogues.[289]

The structural basis of these property differences is stillnot fully established. Because of the fibrils are noncrystalline,insoluble, and mesoscopically heterogeneous nature, NMRspectroscopy rather than X-ray crystallography is the primarytool for fibril structure determination.[285, 290] NMR spectros-copy provides various structural constraints that, whencombined with computational tools, such as geometry opti-mization and MD simulations, yield plausible structuralmodels. The model of Ab42 structure was proposed byRiek et al.[290] and denoted M42. M42 can be dissected intothree motifs; 1) residues 1–16 are randomly coiled, 2) resi-dues 26–31 are the turn and 3) the remaining residues formtwo b strands. NMR structural information is primarilyrelated to the b-strand. As a result of the lack of structuralconstraints, the turn structure in this model is obtained bygeometry optimization and depends heavily on the computa-tional method and the empirical force field. 2D IR spectrawere reported recently.[109,110]

The simulated absorption of M42 (Figure 29 left, panelABS) shows an intense peak at 1635 cm�1 (a), a shoulder at1655 cm�1 (b), a peak at 1675 cm�1 (c), two additional peaks

at 1695 cm�1 (d) and 1715 cm�1 (i). Figure 29 left, panel NMD,shows the normal-mode decomposition of the various normalmodes into the three structural motifs (b-sheet, turn, andcoil). Peaks a, b, and c have strong contributions from both b-strand and coil. Peak d has a contribution from turn plus coil,and peak i is purely turn. Figure 29 left, panel 2D, displays thesimulated xxyy 2DCS signal. The signal is dominated bystrong and broad diagonal peaks that resemble the absorp-tion, no cross-peaks are observed. The contributions of thethree structural motifs overlap. The lower resolution andnormal-mode delocalization complicate the interpretation ofthe cross-peaks compared to the NMR spectra. However,isotope-labeling combined with a judicious design of polar-ization configurations can be used to manipulate the 2DCSsignals by enhancing desired spectral features. 13C18O isotopelabeling of a given peptide residue induces a 65 cm�1 red shiftof the amide-I vibrational frequency, creating peaks wellseparated from the unlabeled band and providing structuralinformation on labeled segments.

2D signals depend on interferences among many contri-butions (Liouville space pathways). This interference may becontrolled by varying the relative polarizations of the variousbeams, thereby eliminating diagonal peaks and amplifying thecross-peaks. Below we demonstrate how a coherent controlalgorithm may be used to manipulate the 2DIR feature ofAb 42, and create well-resolved cross-peaks which are directlyrelated to interactions within turn segments and between theturn and the b-sheet. These provide additional constraints forthe turn structure.

We have optimized the following of the three linearlyindependent tensor components Tj = xxyy ; xyxy ; xyyx tosuppress the diagonal 1655 cm�1 peak [Eq. (42)].

The coefficients cj were optimized using a geneticalgorithm[291] aimed at minimizing the control target: theratio of the integrated diagonal line in the absolute magnitudeof the 2D spectrum to the integrated diagonal peak at1655 cm�1 with d = 10 cm�1. Fast exponential convergencewas achieved using 10 members in a population within 100–200 generations. A much richer cross-peak pattern is seen inthe signal (Figure 29 left, panel 2D(CP)) compared with thenon-controlled xxxx signal (Figure 29 left, panel 2D).

The CP signal of M42 shows two strong cross-peaksrelated to the correlation between the absorption features dand i. These are displayed in Figure 30 on an expanded scaleand marked AB-1 (1695,1715) and AB-2 (1715,1695). Thenormal modes contributing to the diagonal peaks wereprojected onto the local amide modes along the backboneto assign the cross-peaks to positions along the structure. Thei modes (Figure 30 A:1715) are localized within the turnsegment and residue 28 has the largest weight, while thed modes (Figure 30 B:1695) are almost evenly distributedamong the coil and the residues 28–30 of the turn. Given thelarge distance between the coil and the turn (see Figure 30)we expect their interaction to be negligible. We thus conclude

Figure 29. Left: From top to bottom: the noraml mode diagram(NMD), the absorption signal (ABS), the xxyy polarization 2D crossspectrum (2D), and the coherent-control optimized polarization[2D(CP)] 2D cross spectrum of unlabeled M42 amyloid. In NMD, theb-strand, coil, and turn content are shown in red, green, and blue,respectively. Right: same quantities for the isotope-labeled coil fibril.


Chemie



that these two cross-peaks reflect turn–turn interactions,especially within the residues 28–30.

Most peaks in the M42 spectra contain contributions frommore than one structural motif, which will complicate theirassignment. Upon isotope labeling of the coil segment(residues 1–16), the peaks will be dominated by one structuralmotif (Figure 29, right). The new shoulder e in the linearabsorption (Figure 29 right, panel ABS) is dominated by thecoil segment. The components a,b, and c are all dominated bythe sheet and d and i belong to the turn. The 2D spectrum(Figure 29 right, panel 2D) has an improved cross-peakresolution over the unlabeled sample, but the main cross-peak pattern is still unresolved. Our coherent-control methodmay be employed to eliminate the diagonal peak of theisotopically labeled peptide at 1655 cm�1 (Figure 29, right,panel 2D(CP)). Most cross-peaks may now be clearlyassigned.

In Figure 30 the signals C:1715 and D:1695 demonstratethat for the coil-labeled sample, peaks d and i are bothdominated by the turn, the cross-peaks CD1 (1695,1715) andCD2 (1715,1695) are thus related to turn–turn interactions. InFigure 30, signals H:1615 and G:1635 shows that the1615 cm�1 and 1635 cm�1 frequency windows are dominatedby the strand motif. The CH, DH, and CG cross-peaks thusoriginate from interactions between the turn and the sheet

motifs close to the turn segment (mainly residues 24–25 andresidues 32–33). The normal modes in the 1675 cm�1 window(Figure 30 E:1675) are also dominated by the sheet motif, thelocal mode population is non-uniformly distributed andcontains no contribution from mode 25. The CE signal thusoriginates primarily from the interaction between the turnand residue 32. The normal modes 1655 cm�1 (Figure 30F:1655), in contrast, have a significant contribution from boththe sheet and the turn, thus the CF, DF1, DF2 peaks shouldcontain mixed information about turn–turn and turn–sheetinteraction. The additional cross-peaks FF and FH, marked byblack arrows are related to sheet–sheet interactions.

11. Summary and Outlook

The computational arsenal presented herein may bereadily applied to describe non-equilibrium processes, pro-vided they are slower than a typical 2D measurementtimescale (ca. 200 fs). We can then assume that the systemis stationary during the measurement but characterized bytime-dependent parameters related to the process understudy (e.g. protein folding, conformational change, or hydro-gen-bond breaking). 2D spectra could then provide “strobo-scopic” snapshots of these processes. The numerical propa-gation (Figure 20) and the stochastic Liouville equationtechniques are not restricted to this limit and may be usedto describe an arbitrary timescale of the dynamics (fast orslow dynamics compared to the measurement).

It is instructive to point out some fundamental connec-tions between 2D spectroscopy and another rapidly develop-ing field of single-molecule spectroscopy.[292] In the course oftime, each molecule in an ensemble undergoes a stochasticevolution and its properties, for example, frequencies, ori-entations, dipole moments fluctuate through couplings withuncontrollable external “bath” degrees of freedom. Bulkmeasurements probe the ensemble average of these stochastictrajectories. Single-molecule spectroscopy dissects the ensem-ble by “brute force”: observing individual trajectories onemolecule at a time. It thus provides considerably moredetailed information than bulk measurements. Nonlinearspectroscopy accomplishes a similar goal by observing theentire ensemble but at multiple time points. There are manypossible microscopic models with very different types oftrajectories that could yield the same ensemble average at agiven time. The multipoint correlation functions obtained bynonlinear spectroscopy have the capacity to distinguishbetween such models, even though individual trajectoriesare not observed. Consider, for example, a chemicallyreactive AÐB system at equilibrium. If the reaction ratesare slow on the spectroscopic time scale, the absorptionspectrum will be simply given by the weighted average ofspecies A and B; No information about the kinetics isavailable from 1D spectroscopy. In a 2D measurement thetime delay t2 can be varied on the kinetic timescale so as toextract the kinetics from the time evolution of the cross-peaks. The cross-peaks give the joint probability of the systemto be in A during t1 and B during t3. Typically t1 and t3 arecontrolled by dephasing and are much shorter than t2. This is

Figure 30. Above the dash line: The 2DCS signal of M42 withcoherent-control-optimized polarization configuration (Figure 29,bottom left panel) on an expanded scale (1630–1730 cm�1) and theprojection of the normal modes contributing to the specified cross-peaks onto the local amide modes along the backbone. Below thedashed line: Same representation for isotope-labeled coil M42. In the2DCS plot, the cross-peaks are attributed to the turn–turn interac-tion (blue arrows), turn–sheet (red), and sheet–sheet (black). In thenormal mode projection plots, the contribution from the turn (blue),sheet (red), and coil (green) are shown. Green arrows above the 2DCSdenote the positions of absorption maxima.




therefore a two-point measurement separated by t2. This iscomplementary to triggered experiments in which the systemis perturbed out of equilibrium and the subsequent relaxationis monitored.[92, 293] Single-molecule spectroscopy is a long(microsecond and longer) time-measurement. 2DIR canprovide trajectory information on the femtosecond timescale.A common thread to both techniques is the analysis in termsof ensembles of trajectories rather than of configurations.[294]

Over the past decade, 2DIR has established itself as auseful spectroscopic tool for the investigation of molecularstructures and ultrafast molecular events. The technique has alower structural resolution than NMR spectroscopy, but itsunique high temporal resolution and different observationwindow make it an invaluable complementary tool to NMRspectroscopy.

Early studies were of the proof-of-concept kind andfocused on demonstrating the various capabilities and poten-tials of this technique. Current activity in the field focuses onidentifying specific systems where 2DIR can be particularlyhelpful. Developing the necessary methods for quantitativelyanalyzing the 2DIR signals is a major challenge. A concertedexperimental and the theoretical effort, will be required tobenchmark systems and improve the current methods. Weexpect it to go through a similar development trajectory to thehistory of classical force field for molecular mechanicsimulations.

A computational package “SPECTRON” has been devel-oped[55] for simulating 2D signals. We aim at calculating abroad range of linear and nonlinear optical signals of complexbiomolecules. SPECTRON includes modules for constructingHamiltonian operators for 1) the amide-I, -II, -III, -A vibra-tional bands and n-p*, p–p* electronic bands for peptides,based on simulations of the MD trajectories, 2) the C=Ostretch in guanine, the in-plane or out-of-plane, symmetric orasymmetric NH or NH2 bend in adenine, the ring C=N stretchin cytosine for RNAs, 3) the O�H stretching band of water,4) C�O stretching band of membrane lipids. The code wasused recently to benchmark various amide maps.[295] Aninterface between SPECTRON and standard MD simulationpackages, such as CHARMM,[58] NAMD,[296] andGROMOS,[59] allows the MD trajectories to be read directlyin ASCII or binary formats. The code can also simulate 2Delectronic spectra of aggregates. This application has beenreviewed recently[64] and is not covered in this Review.

This work was supported by the National Institutes of HealthGrant GM59230 and the National Science Foundation GrantCHE-0745891. W.Z. thanks UCI Dissertation Fellowship forfinancial support. Many helpful discussions with Dr. DariusAbramavicius are gratefully acknowledged. We also wish tothank Drs. Cyril Falvo and Lijun Yang for useful comments.

Received: June 5, 2008Revised: September 17, 2008

[1] L. Stryer, Biochemistry, 2nd ed, Freeman, New York, 1995.[2] G. Rhodes, Crystallography Made Crystal Clear: A Guide for

Users of Macromolecular Models, 3rd ed., Academic Press,Burlington, 2006.

[3] D. J. Segel, A. Bachmann, J. Hofrichter, K. O. Hodgson, S.Doniach, T. Kiefhaber, J. Mol. Biol. 1999, 288, 489.

[4] S. Arai, M. Hirai, Biophys. J. 1999, 76, 2192 – 2197.[5] L. Pollack, M. W. Tate, N. C. Darnton, J. B. Knight, S. M.

Gruner, W. A. Eaton, R. H. Austin, Proc. Natl. Acad. Sci. USA1999, 96, 10115 – 10117.

[6] T. Uzawa, T. Kimura, K. Ishimori, I. Morishima, T. Matsui, M.Ikeda-Saito, S. Takahashi, S. Akiyama, T. Fujisawa, J. Mol. Biol.2006, 357, 997 – 1008.

[7] M. Pfuhl, P. C. Driscoll, Philos. Trans. R. Soc. London Ser. A2000, 358, 513 – 545.

[8] K. W�thrich, NMR of Proteins and Nucleic Acids, Wiley, NewYork, 1995.

[9] B. Gruenewald, C. U. Nicola, A. Lustig, G. Schwarz, H. Klump,Biophys. Chem. 1979, 9, 137 – 147.

[10] R. R. Ernst, G. Bodenhausen, A. Wokaun, Principles ofNuclear Magnetic Resonance in One and Two Dimensions,Clarendon, Oxford, 1987.

[11] C. Rose-Petruck, R. Jimenez, T. Guo, A. Cavalleri, C. W.Siders, F. Raksi, J. A. Squier, B. C. Walker, K. R. Wilson, C. P. J.Barty, Nature 1999, 398, 310 – 312.

[12] M. Chergui, A. H. Zewail, ChemPhysChem, 2009, 10, 28 – 43.[13] A. H. Zewail, Physical Biology, From Atoms to Medicine,

Imperial College Press, London, 2008.[14] W. A. Eaton, V. Munoz, S. J. Hagen, G. S. Jas, L. J. Lapidus,

E. R. Henry, J. Hofrichter, Annu. Rev. Biophys. Biomol. Struct.2000, 29, 327 – 359.

[15] Applications of Vibrational Spectroscopy in Life, Pharmaceut-ical and Natural Sciences (Eds.: J. M. Chalmers, P. R. Griffiths),Wiley, New York, 2002.

[16] M. Hein, A. A. Wegener, M. Engelhard, F. Siebert, Biophys. J.2003, 84, 1208 – 1217.

[17] R. Brudler, R. Rammelsberg, T. T. Woo, E. D. Getzor, K.Gerwert, Nat. Struct. Biol. 2001, 8, 265 – 270.

[18] M. Aki, T. Ogura, K. Shinzawa-Itoh, S. Yoshikawa, T.Kitagawa, J. Phys. Chem. B 2000, 104, 10765 – 10774.

[19] S. A. Asher, A. Ianoul, G. Mix, M. N. Boyden, A. Karnoup, M.Diem, R. Schweitzer-Stenner, J. Am. Chem. Soc. 2001, 123,11775 – 11781.

[20] E. W. Blanch, L. A. Morozova-Roche, D. A. E. Cochran, A. J.Doig, L. Hecht, L. D. Barron, J. Mol. Biol. 2000, 301, 553 – 563.

[21] M. H. Cho, Chem. Rev. 2008, 108, 1331 – 1418.[22] J. Bredenbeck, J. Helbing, C. Kolano, P. Hamm, ChemPhys-

Chem 2007, 8, 1747 – 1756.[23] C. Scheurer, S. Mukamel, J. Chem. Phys. 2001, 115, 4989.[24] C. Scheurer, S. Mukamel, J. Chem. Phys. 2002, 116, 6803 – 6816.[25] C. Scheurer, S. Mukamel, Bull. Chem. Soc. Jpn. 2002, 75, 989 –

999.[26] K. W. Kwak, S. Park, M. D. Fayer, Proc. Natl. Acad. Sci. USA

2007, 104, 14221 – 14226.[27] R. Venkatramani, S. Mukamel, J. Chem. Phys. 2002, 117,

11089 – 11101.[28] Chem. Phys. 2007, 341(1–3) (special issue: “Ultrafast Dynamics

of Molecules in the Condensed Phase”).[29] W. Zhao, J. C. Wright, J. Am. Chem. Soc. 1999, 121, 10994 –

10998.[30] E. I. Shakhnovich, Curr. Opin. Struct. Biol. 1997, 7, 29 – 40.[31] M. Karplus, J. A. McCammon, Nat. Struct. Biol. 2002, 9, 646 –

652.[32] J. A. McCammon, S. C. Harvey, Dynamics of Proteins and

Nucleic Acids, Cambridge University Press, Cambridge, 1987.[33] X. Daura, K. Gademann, H. Sch�fer, B. Jaun, D. Seebach, W. F.

van Gunsteren, J. Am. Chem. Soc. 2001, 123, 2393 – 2404.[34] J. N. Onuchic, Z. Luthey-Schulten, P. G. Wolynes, Annu. Rev.

Phys. Chem. 1997, 48, 545 – 600.[35] J. D. Bryngelson, J. N. Onuchic, N. D. Socci, P. G. Wolynes,

Proteins Struct. Funct. Genet. 1995, 21, 167 – 195.


Chemie


http://dx.doi.org/10.1006/jmbi.1999.2703

http://dx.doi.org/10.1016/S0006-3495(99)77374-1

http://dx.doi.org/10.1073/pnas.96.18.10115


http://dx.doi.org/10.1016/j.jmb.2005.12.089

http://dx.doi.org/10.1016/j.jmb.2005.12.089

http://dx.doi.org/10.1016/0301-4622(79)87008-8

http://dx.doi.org/10.1038/18631

http://dx.doi.org/10.1146/annurev.biophys.29.1.327

http://dx.doi.org/10.1146/annurev.biophys.29.1.327

http://dx.doi.org/10.1016/S0006-3495(03)74935-2

http://dx.doi.org/10.1016/S0006-3495(03)74935-2

http://dx.doi.org/10.1038/85021

http://dx.doi.org/10.1021/jp000357p

http://dx.doi.org/10.1021/ja0039738



http://dx.doi.org/10.1021/cr078377b

http://dx.doi.org/10.1002/cphc.200700148

http://dx.doi.org/10.1002/cphc.200700148

http://dx.doi.org/10.1063/1.1391266

http://dx.doi.org/10.1063/1.1461362

http://dx.doi.org/10.1246/bcsj.75.989


http://dx.doi.org/10.1073/pnas.0701710104


http://dx.doi.org/10.1063/1.1518001

http://dx.doi.org/10.1063/1.1518001



http://dx.doi.org/10.1016/S0959-440X(97)80005-X

http://dx.doi.org/10.1038/nsb0902-646

http://dx.doi.org/10.1038/nsb0902-646

http://dx.doi.org/10.1021/ja003689g

http://dx.doi.org/10.1146/annurev.physchem.48.1.545



[36] P. G. Wolynes, J. N. Onuchic, D. Thirumalai, Science 1995, 267,1619 – 1620.

[37] S. Gnanakaran, H. Nymeyer, J. Portman, K. Y. Sanbonmatsu,A. E. Garcia, Curr. Opin. Struct. Biol. 2003, 13, 168 – 174.

[38] Y. Duan, P. Kollman, Science 1998, 282, 740.[39] J. Wang, J. Onuchic, P. Wolynes, Phys. Rev. Lett. 1996, 76, 4861 –

4864.[40] C. D. Snow, B. Zagrovic, V. S. Pande, J. Am. Chem. Soc. 2002,

124, 14548 – 14549.[41] P. L. Freddolino, F. Liu, M. Gruebele, K. Schulten, Biophys. J.

2008, 94, L75 – L77.[42] Y. Zhu, D. O. V. Alonso, K. Maki, C. Y. Huang, S. J. Lahr, V.

Daggett, H. Roder, W. F. DeGrado, F. Gai, Proc. Natl. Acad.Sci. USA 2003, 100, 15486 – 15491.

[43] J. Kubelka, W. A. Eaton, J. Hofrichter, J. Mol. Biol. 2003, 329,625 – 630.

[44] C. D. Snow, N. Nguyen, V. S. Pande, M. Gruebele, Nature 2002,420, 102 – 106.

[45] D. Frenkel, B. Smit, Understanding Molecular Simulation:From Algorithms to Applications, Academic Press, San Diego,2002.

[46] D. Mohanty, R. Elber, D. Thirumalai, D. Beglov, B. Roux, J.Mol. Biol. 1997, 272, 423 – 442.

[47] C. Dellago, P. Bolhuis, F. Csajka, D. Chandler, J. Chem. Phys.1998, 108, 1964 – 1977.

[48] V. Chernyak, W. M. Zhang, S. Mukamel, J. Chem. Phys. 1998,109, 9587 – 9601.

[49] S. Mukamel, A. Piryatinski, V. Chernyak, J. Chem. Phys. 1999,110, 1711 – 1725.

[50] A. Piryatinski, S. Tretiak, V. Chernyak, S. Mukamel, J. RamanSpectrosc. 2000, 31, 125 – 135.

[51] A. Piryatinski, V. Chernyak, S. Mukamel in Ultrafast Infraredand Raman Spectroscopy (Ed.: M. Fayer), Marcel Dekker, NewYork, 2001, pp. 349 – 382.

[52] A. Piryatinski, V. Chernyak, S. Mukamel, Chem. Phys. 2001,266, 311 – 322.

[53] C. Scheurer, A. Piryatinski, S. Mukamel, J. Am. Chem. Soc.2001, 123, 3114 – 3124.

[54] A. Piryatinski, S. A. Asher, S. Mukamel, J. Phys. Chem. A 2002,106, 3524 – 3530.

[55] W. Zhuang, D. Abramavicius, T. Hayashi, S. Mukamel, J. Phys.Chem. B 2006, 108, 18 034 – 18045.

[56] J. P. Wang, J. X. Chen, R. M. Hochstrasser, J. Phys. Chem. B2006, 110, 7545 – 7555.

[57] N. Demirdoven, C. M. Cheatum, H. S. Chung, M. Khalil, J.Knoester, A. Tokmakoff, J. Am. Chem. Soc. 2004, 126, 7981 –7990.

[58] B. R. Brooks, R. E. Bruccoleri, B. D. Olafson, D. J. States, S.Swaminathan, M. Karplus, J. Comput. Chem. 1983, 4, 187 – 217.

[59] W. R. P. Scott, P. H. H�nenberger, I. G. Tironi, A. E. Mark,S. R. Billeter, J. Fennen, A. E. Torda, T. Huber, P. Kr�ger, W. F.van Gunsteren, J. Phys. Chem. A 1999, 103, 3596 – 3607.

[60] D. A. Pearlman, D. A. Case, J. W. Caldwell, W. S. Ross, T. E.Cheatham, S. Debolt, D. Ferguson, G. Seibel, P. Kollman,Comput. Phys. Commun. 1995, 91, 1 – 41.

[61] D. Abramavicius, S. Mukamel, J. Chem. Phys. 2005, 122,134305.

[62] S. Mukamel, D. Abramavicius, Chem. Rev. 2004, 104, 2073 –2098.

[63] S. Mukamel in Molecular Nonlinear Optics (Ed.: J. Zyss),Academic Press, New York, 1994, pp. 1 – 46.

[64] D. Abramavicius, B. Palmieri, V. Voronine, F. Sanda, S.Mukamel, Chem. Rev. 2008, in press.

[65] S. Mukamel, Annu. Rev. Phys. Chem. 2000, 51, 691 – 729.[66] S. Mukamel, Principles of Nonlinear Optical Spectroscopy,

Oxford University Press, New York, 1995.

[67] S. M. George, A. L. Harris, M. Berg, C. B. Harris, J. Chem.Phys. 1984, 80, 83 – 94.

[68] L. Allen, J. H. Elberly, Optical Resonances and Two-LevelAtoms, Wiley, New York, 1975.

[69] A. Laubereau, W. Kaiser, Rev. Mod. Phys. 1978, 50, 607 – 665.[70] R. F. Loring, S. Mukamel, J. Chem. Phys. 1985, 83, 2116 – 2128.[71] L. J. Muller, D. Vandenbout, M. Berg, J. Chem. Phys. 1993, 99,

810 – 819.[72] D. Vandenbout, L. J. Muller, M. Berg, Phys. Rev. Lett. 1991, 67,

3700 – 3703.[73] M. Muller, K. Wynne, J. D. W. Vanvoorst, Chem. Phys. 1988,

125, 225 – 230.[74] S. Mukamel, Phys. Rev. A 1983, 28, 3480 – 3492.[75] S. Mukamel, R. F. Loring, J. Opt. Soc. Am. B 1986, 3, 595 – 606.[76] L. E. Fried, S. Mukamel, Adv. Chem. Phys. 1993, 84, 435.[77] Y. Tanimura, S. Mukamel, J. Chem. Phys. 1993, 99, 9496 – 9511.[78] L. J. Kaufman, J. Heo, L. D. Ziegler, G. R. Fleming, Phys. Rev.

Lett. 2002, 88, 207402.[79] K. Kubarych, C. J. Milne, R. J. D. Miller, Chem. Phys. Lett.

2003, 369, 635 – 642.[80] S. Saito, I. Ohmine, J. Chem. Phys. 1998, 108, 250 – 251.[81] A. Ma, R. M. Stratt, J. Chem. Phys. 2002, 116, 4972 – 4984.[82] K. Okumura, A. Tokmakoff, Y. Tanimura, J. Chem. Phys. 1999,

111, 492.[83] S. Mukamel, A. Piryatinski, V. Chernyak, Acc. Chem. Res. 1999,

32, 145 – 154.[84] Ultrafast Phenomena XIII (Eds.: R. D. Miller, M. M. Muri-

name, N. F. Scherer, A. M. Weiner), Springer, Heidelberg, 2002.[85] T. L. C. Jansen, K. Duppen, J. G. Snijders, Phys. Rev. B 2003, 67,

134206.[86] S. Saito, I. Ohmine, J. Chem. Phys. 2003, 119, 9073 – 9087.[87] C. J. Milne, Y. Li, R. J. D. Miller in Time-Resolved Spectroscopy

in Complex Liquids (Ed.: R. Torre), Springer, New York, 2007,p. 1.

[88] S. Palese, J. T. Buontempo, L. Schilling, W. T. Lotshaw, Y.Tanimura, S. Mukamel, R. J. D. Miller, J. Phys. Chem. 1994, 98,12466 – 12470.

[89] Y. L. Li, L. Huang, R. J. D. Miller, T. Hasegawa, Y. Tanimura, J.Chem. Phys. 2008, 128, 234507.

[90] W. M. Zhang, V. Chernyak, S. Mukamel, J. Chem. Phys. 1999,110, 5011.

[91] M. C. Asplund, M. T. Zanni, R. M. Hochstrasser, Proc. Natl.Acad. Sci. USA 2000, 97, 8219 – 8224.

[92] C. Kolano, J. Helbing, M. Kozinski, W. Sander, P. Hamm,Nature 2006, 444, 469 – 472.

[93] P. Mukherjee, I. Kass, I. Arkin, M. T. Zanni, Proc. Natl. Acad.Sci. USA 2006, 103, 3528 – 3533.

[94] T. Brixner, J. Stenger, H. M. Vaswani, M. Cho, R. E. Blanken-ship, G. R. Fleming, Nature 2005, 434, 625 – 628.

[95] J. D. Hybl, Y. Christophe, D. M. Jonas, Chem. Phys. 2001, 266,295 – 309.

[96] X. Q. Li, T. H. Zhang, C. N. Borca, S. T. Cundiff, Phys. Rev.Lett. 2006, 96, 57406.

[97] T. H. Zhang, I. Kuznetsova, T. Meier, X. C. Li, R. P. Mirin, P.Thomas, S. T. Cundiff, Proc. Natl. Acad. Sci. USA 2007, 104,14227 – 14232.

[98] P. Hamm, M. H. Lim, R. M. Hochstrasser, J. Phys. Chem. B1998, 102, 6123 – 6138.

[99] P. Hamm, M. Lim, W. F. DeGrado, R. M. Hochstrasser, Proc.Natl. Acad. Sci. USA 1999, 96, 2036 – 2041.

[100] S. Woutersen, P. Hamm, J. Phys. Chem. B 2000, 104, 11316 –11320.

[101] M. T. Zanni, M. C. Asplund, R. M. Hochstrasser, J. Chem.Phys. 2001, 114, 4579 – 4590.

[102] C. Fang, J. Wang, Y. S. Kim, A. K. Charnley, W. Barber-Armstrong, A. B. Smith, S. M. Decatur, R. M. Hochstrasser, J.Phys. Chem. B 2004, 108, 10415 – 10427.



http://dx.doi.org/10.1126/science.7886447


http://dx.doi.org/10.1016/S0959-440X(03)00040-X

http://dx.doi.org/10.1126/science.282.5389.740

http://dx.doi.org/10.1103/PhysRevLett.76.4861


http://dx.doi.org/10.1021/ja028604l

http://dx.doi.org/10.1021/ja028604l

http://dx.doi.org/10.1529/biophysj.108.131565

http://dx.doi.org/10.1529/biophysj.108.131565



http://dx.doi.org/10.1016/S0022-2836(03)00519-9

http://dx.doi.org/10.1016/S0022-2836(03)00519-9

http://dx.doi.org/10.1038/nature01160




http://dx.doi.org/10.1063/1.475562

http://dx.doi.org/10.1063/1.475562

http://dx.doi.org/10.1063/1.477621

http://dx.doi.org/10.1063/1.477621

http://dx.doi.org/10.1063/1.478030

http://dx.doi.org/10.1063/1.478030

http://dx.doi.org/10.1002/(SICI)1097-4555(200001/02)31:1/2%3C125::AID-JRS503%3E3.0.CO;2-O

http://dx.doi.org/10.1002/(SICI)1097-4555(200001/02)31:1/2%3C125::AID-JRS503%3E3.0.CO;2-O

http://dx.doi.org/10.1016/S0301-0104(01)00253-1

http://dx.doi.org/10.1016/S0301-0104(01)00253-1



http://dx.doi.org/10.1021/jp012849n

http://dx.doi.org/10.1021/jp012849n

http://dx.doi.org/10.1021/jp057564f


http://dx.doi.org/10.1002/jcc.540040211


http://dx.doi.org/10.1016/0010-4655(95)00041-D

http://dx.doi.org/10.1063/1.1869495

http://dx.doi.org/10.1063/1.1869495




http://dx.doi.org/10.1063/1.446410

http://dx.doi.org/10.1063/1.446410

http://dx.doi.org/10.1103/RevModPhys.50.607

http://dx.doi.org/10.1063/1.449302

http://dx.doi.org/10.1063/1.465344

http://dx.doi.org/10.1063/1.465344

http://dx.doi.org/10.1103/PhysRevA.28.3480

http://dx.doi.org/10.1364/JOSAB.3.000595

http://dx.doi.org/10.1002/9780470141427.ch6

http://dx.doi.org/10.1063/1.465484



http://dx.doi.org/10.1016/S0009-2614(03)00039-3

http://dx.doi.org/10.1016/S0009-2614(03)00039-3

http://dx.doi.org/10.1063/1.1453402

http://dx.doi.org/10.1063/1.479383

http://dx.doi.org/10.1063/1.479383

http://dx.doi.org/10.1021/ar960206y

http://dx.doi.org/10.1021/ar960206y

http://dx.doi.org/10.1103/PhysRevB.67.134206

http://dx.doi.org/10.1103/PhysRevB.67.134206

http://dx.doi.org/10.1063/1.1609984

http://dx.doi.org/10.1021/j100099a003

http://dx.doi.org/10.1021/j100099a003

http://dx.doi.org/10.1063/1.478400

http://dx.doi.org/10.1063/1.478400







http://dx.doi.org/10.1016/S0301-0104(01)00233-6

http://dx.doi.org/10.1016/S0301-0104(01)00233-6



http://dx.doi.org/10.1021/jp9813286




http://dx.doi.org/10.1021/jp001546a


http://dx.doi.org/10.1063/1.1346647

http://dx.doi.org/10.1063/1.1346647

http://dx.doi.org/10.1021/jp049893y



[103] H. Maekawa, C. Toniolo, A. Moretto, Q. B. Broxterman, N. H.Ge, J. Phys. Chem. B 2006, 110, 5834 – 5837.

[104] A. T. Krummel, P. Mukherjee, M. T. Zanni, J. Phys. Chem. B2003, 107, 9165 – 9169.

[105] O. F. A. Larsen, P. Bodis, W. J. Buma, J. S. Hannam, D. A.Leigh, S. Woutersen, Proc. Natl. Acad. Sci. USA 2005, 102,13378 – 13382.

[106] Z. Ganim, H. S. Chung, A. W. Smith, L. P. Deflores, K. C. Jones,A. Tokmakoff, Acc. Chem. Res. 2008, 41, 432 – 441.

[107] V. Volkov, R. Chelli, W. Zhuang, F. Nuti, Y. Takaoka, A. M.Papini, S. Mukamel, R. Righini, Proc. Natl. Acad. Sci. USA2007, 104, 15323 – 15327.

[108] J. Bredenbeck, A. Ghosh, M. Smits, M. Bonn, J. Am. Chem.Soc. 2008, 130, 2152 – 2153.

[109] S. H. Shim, D. B. Strasfeld, Y. L. Ling, M. T. Zanni, Proc. Natl.Acad. Sci. USA 2007, 104, 14197 – 14202.

[110] C. Paul, J. P. Wang, W. C. Wimley, R. M. Hochstrasser, P. H.Axelsen, J. Am. Chem. Soc. 2004, 126, 5843 – 5850.

[111] Y. S. Kim, L. Liu, P. H. Axelsen, R. M. Hochstrasser, Proc. Natl.Acad. Sci. USA 2008, 105, 7720 – 7725.

[112] J. Bredenbeck, J. Helbing, R. Behrendt, C. Renner, L. Moroder,J. Wachtveitl, P. Hamm, J. Phys. Chem. B 2003, 107, 8654 – 8660.

[113] H. S. Chung, Z. Ganim, K. C. Jones, A. Tokmakoff, Proc. Natl.Acad. Sci. USA 2007, 104, 14237 – 14242.

[114] J. Bredenbeck, J. Helbing, J. R. Kumita, G. A. Woolley, P.Hamm, Proc. Natl. Acad. Sci. USA 2005, 102, 2379 – 2384.

[115] M. Broquier, C. Crepin, H. Dubost, J. P. Galaup, Chem. Phys.2007, 341, 207 – 217.

[116] J. N. Bandaria, S. Dutta, S. E. Hill, A. Kohen, C. M. Cheatum, J.Am. Chem. Soc. 2008, 130, 22 – 23.

[117] J. F. Cahoon, K. R. Sawyer, J. P. Schlegel, C. B. Harris, Science2008, 319, 1820 – 1823.

[118] J. C. Deak, S. T. Rhea, L. K. Iwaki, D. D. Dlott, J. Phys. Chem.A 2000, 104, 4866 – 4875.

[119] D. D. Dlott, Chem. Phys. 2001, 266, 149 – 166.[120] C. J. Fecko, J. D. Eaves, J. J. Loparo, A. Tokmakoff, P. L.

Geissler, Science 2003, 301, 1698 – 1702.[121] J. D. Eaves, J. J. Loparo, C. J. Fecko, S. T. Roberts, A. Tokmak-

off, P. L. Geissler, Proc. Natl. Acad. Sci. USA 2005, 102, 13019 –13022.

[122] J. B. Asbury, T. Steinel, C. Stromberg, S. A. Corcelli, C. P.Lawrence, J. L. Skinner, M. D. Fayer, J. Phys. Chem. A 2004,108, 1107 – 1119.

[123] D. Kraemer, M. L. Cowan, A. Paarmann, N. Huse, E. T. J.Nibbering, T. Elsaesser, R. J. D. Miller, Proc. Natl. Acad. Sci.USA 2008, 105, 437 – 442.

[124] P. Hamm, M. Lim, R. M. Hochstrasser, Phys. Rev. Lett. 1998,81, 5326 – 5329.

[125] M. Kozinski, S. Garrett-Roe, P. Hamm, Chem. Phys. 2007, 341,5 – 10.

[126] C. H. Kuo, R. M. Hochstrasser, Chem. Phys. 2007, 341, 21 – 28.[127] J. J. Sines, S. A. Allison, J. A. McCammon, Biochemistry 1990,

29, 9403 – 9412.[128] D. K. Jones-Hertzog, W. L. Jorgensen, J. Med. Chem. 1997, 40,

1539 – 1549.[129] B. E. Cohen, T. B. McAnaney, E. S. Park, Y. N. Jan, S. G. Boxer,

L. Y. Jan, Science 2002, 296, 1700 – 1703.[130] C. Fang, J. D. Bauman, K. Das, A. Remorino, E. Arnold, R. M.

Hochstrasser, Proc. Natl. Acad. Sci. USA 2008, 105, 1472 – 1477.[131] I. T. Suydam, S. G. Boxer, Biochemistry 2003, 42, 12050 – 12055.[132] A. Piryatinski, V. Chernyak, S. Mukamel, Chem. Phys. 2001,

266, 285 – 294.[133] D. V. Kurochkin, S. R. G. Naraharisetty, I. V. Rubtsov, Proc.

Natl. Acad. Sci. USA 2007, 104, 14209 – 14214.[134] Q. Shi, E. Geva, J. Phys. Chem. A 2003, 107, 9059 – 9069.[135] Y. S. Kim, R. M. Hochstrasser, Proc. Natl. Acad. Sci. USA 2005,

102, 11185 – 11190.

[136] J. R. Zheng, K. Kwak, J. Asbury, X. Chen, I. R. Piletic, M. D.Fayer, Science 2005, 309, 1338 – 1343.

[137] J. Bredenbeck, J. Helbing, K. Nienhaus, G. U. Nienhaus, P.Hamm, Proc. Natl. Acad. Sci. USA 2007, 104, 14243 – 14248.

[138] O. Golonzka, M. Khalil, N. Demirdoven, A. Tokmakoff, J.Chem. Phys. 2001, 115, 10814 – 10828.

[139] T. Miyazawa, T. Shimanouchi, S. I. Mizushima, J. Chem. Phys.1958, 29, 611 – 616.

[140] M. Tsuboi, T. Onishi, I. Nakagawa, T. Shimanouchi, S.Mizushima, Spectrochim. Acta 1958, 12, 253 – 261.

[141] S. Krimm, J. Bandekar, Adv. Protein Chem. 1986, 38, 181 – 364.[142] T. Hayashi, W. Zhuang, S. Mukamel, J. Phys. Chem. A 2005,

109, 9747 – 9759.[143] J. T. Pelton, L. R. McLean, Anal. Biochem. 2000, 277, 167 – 176.[144] Infrared Spectroscopy of Biomolecules (Eds.: H. H. Mantsch,

D. Chapman), Wiley-Liss, New York, 1996.[145] T. Wang, Y. J. Zhu, Z. Getahun, D. G. Du, C. Y. Huang, W. F.

DeGrado, F. Gai, J. Phys. Chem. B 2004, 108, 15301 – 15310.[146] H. Torii, M. Tasumi, J. Raman Spectrosc. 1998, 29, 81 – 86.[147] See Ref. [142].[148] T. Hayashi, S. Mukamel, J. Phys. Chem. B 2007, 111, 11032 –

11046.[149] see Ref. [141].[150] H. Torii, M. Tasumi, J. Chem. Phys. 1992, 96, 3379 – 3387.[151] R. D. Gorbunov, D. S. Kosov, G. Stock, J. Chem. Phys. 2005,

122, 224904.[152] P. Hamm, S. Woutersen, Bull. Chem. Soc. Jpn. 2002, 75, 985 –

988.[153] T. L. Jansen, A. G. Dijkstra, T. M. Watson, J. D. Hirst, J.

Knoester, J. Chem. Phys. 2006, 125, 044312.[154] J. R. Cheeseman, M. J. Frisch, F. J. Devlin, P. J. Stephens, Chem.

Phys. Lett. 1996, 252, 211 – 220.[155] T. Hayashi, H. Hamaguchi, Chem. Phys. Lett. 2000, 326, 115 –

122.[156] K. A. Merchant, W. G. Noid, R. Akiyama, I. J. Finkelstein, A.

Goun, B. L. McClain, R. F. Loring, M. D. Fayer, J. Am. Chem.Soc. 2003, 125, 13804 – 13818.

[157] J. R. Hill, A. Tokmakoff, K. A. Peterson, B. Sauter, D. Zimdars,D. D. Dlott, M. D. Fayer, J. Phys. Chem. 1994, 98, 11213 – 11219.

[158] S. Ham, J. H. Kim, H. Lee, M. H. Cho, J. Chem. Phys. 2003, 118,3491 – 3498.

[159] K. Kwac, M. H. Cho, J. Chem. Phys. 2003, 119, 2247 – 2255.[160] J. R. Schmidt, S. A. Corcelli, J. L. Skinner, J. Chem. Phys. 2004,

121, 8887 – 8896.[161] T. M. Watson, J. D. Hirst, Mol. Phys. 2005, 103, 1531 – 1546.[162] T. Hayashi, S. Mukamel, J. Chem. Phys. 2006, 125, 194510.[163] T. L. Jansen, J. Knoester, J. Chem. Phys. 2006, 124, 044502.[164] W. Zhuang, D. Abramavicius, T. Hayashi, S. Mukamel, J. Phys.

Chem. B 2006, 110, 3362 – 3374.[165] B. R. Brooks, R. E. Bruccoleri, B. D. Olafson, D. J. States, S.

Swaminathan, M. Karplus, J. Comput. Chem. 1983, 4, 187 – 217.[166] Gaussian 03 (Revision c.01), Technical report, M. J. Frisch

et al., 2003.[167] J. Kubelka, T. A. Keiderling, J. Phys. Chem. A 2001, 105,

10922 – 10928.[168] See Ref. [166].[169] T. Hayashi, S. Mukamel, J. Phys. Chem. A 2003, 107, 9113 –

9131.[170] T. Hayashi, S. Mukamel, Bull. Korean Chem. Soc. 2003, 24,

1097 – 1101.[171] L. C. Mayne, B. Hudson, J. Phys. Chem. 1991, 95, 2962 – 2967.[172] H. Guo, M. Karplus, J. Phys. Chem. 1992, 96, 7273 – 7287.[173] S. Ham, J. H. Kim, H. Lee, M. H. Cho, J. Chem. Phys. 2003, 118,

3491 – 3498.[174] D. Thouless, Phys. Rep. 1974, 13, 94 – 142.


Chemie


http://dx.doi.org/10.1021/jp057472q

http://dx.doi.org/10.1021/jp035473h

http://dx.doi.org/10.1021/jp035473h



http://dx.doi.org/10.1021/ar700188n



http://dx.doi.org/10.1021/ja710099c

http://dx.doi.org/10.1021/ja710099c



http://dx.doi.org/10.1021/ja038869f



http://dx.doi.org/10.1021/jp034552q




http://dx.doi.org/10.1016/j.chemphys.2007.06.030


http://dx.doi.org/10.1021/ja077599o

http://dx.doi.org/10.1021/ja077599o



http://dx.doi.org/10.1016/S0301-0104(01)00225-7




http://dx.doi.org/10.1021/jp036266k







http://dx.doi.org/10.1021/bi00492a014


http://dx.doi.org/10.1021/jm960684e

http://dx.doi.org/10.1021/jm960684e



http://dx.doi.org/10.1021/bi0352926

http://dx.doi.org/10.1016/S0301-0104(01)00231-2

http://dx.doi.org/10.1016/S0301-0104(01)00231-2



http://dx.doi.org/10.1021/jp030497+





http://dx.doi.org/10.1063/1.1417504

http://dx.doi.org/10.1063/1.1417504

http://dx.doi.org/10.1063/1.1744547

http://dx.doi.org/10.1063/1.1744547

http://dx.doi.org/10.1016/0371-1951(58)80041-7

http://dx.doi.org/10.1016/S0065-3233(08)60528-8

http://dx.doi.org/10.1021/jp052324l

http://dx.doi.org/10.1021/jp052324l

http://dx.doi.org/10.1006/abio.1999.4320

http://dx.doi.org/10.1021/jp037272j

http://dx.doi.org/10.1002/(SICI)1097-4555(199801)29:1%3C81::AID-JRS214%3E3.0.CO;2-H

http://dx.doi.org/10.1021/jp070369b

http://dx.doi.org/10.1021/jp070369b

http://dx.doi.org/10.1063/1.461939

http://dx.doi.org/10.1063/1.1898215

http://dx.doi.org/10.1063/1.1898215



http://dx.doi.org/10.1016/0009-2614(96)00154-6

http://dx.doi.org/10.1016/0009-2614(96)00154-6

http://dx.doi.org/10.1016/S0009-2614(00)00755-7

http://dx.doi.org/10.1016/S0009-2614(00)00755-7

http://dx.doi.org/10.1021/ja035654x

http://dx.doi.org/10.1021/ja035654x

http://dx.doi.org/10.1021/j100094a032

http://dx.doi.org/10.1063/1.1536980

http://dx.doi.org/10.1063/1.1536980

http://dx.doi.org/10.1063/1.1580807

http://dx.doi.org/10.1063/1.1791632

http://dx.doi.org/10.1063/1.1791632

http://dx.doi.org/10.1080/00268970500052387

http://dx.doi.org/10.1063/1.2348865

http://dx.doi.org/10.1021/jp055813u

http://dx.doi.org/10.1021/jp055813u




http://dx.doi.org/10.1021/jp030626m

http://dx.doi.org/10.1021/jp030626m

http://dx.doi.org/10.1021/j100161a006

http://dx.doi.org/10.1021/j100197a027

http://dx.doi.org/10.1063/1.1536980

http://dx.doi.org/10.1063/1.1536980

http://dx.doi.org/10.1016/0370-1573(74)90029-5


[175] M. L. Cowan, B. D. Bruner, N. Huse, J. R. Dwyer, B. Chugh,E. T. J. Nibbering, T. Elsaesser, R. J. D. Miller, Nature 2005,434, 199 – 202.

[176] I. V. Schweigert, S. Mukamel, Phys. Rev. A 2008, 77, 033802.[177] R. A. Marcus, Adv. Chem. Phys. 2007, 106, 1 – 6.[178] C. Fang, J. Wang, Y. S. Kim, A. K. Charnley, W. Barber-

Armstrong, A. B. Smith, S. M. Decatur, R. M. Hochstrasser, J.Phys. Chem. B 2004, 108, 10415 – 10427.

[179] M. Kobus, R. D. Gorbunov, P. H. Nguyen, G. Stock, Chem.Phys. 2008, 347, 208 – 217.

[180] R. Kubo, J. Math. Phys. 1963, 4, 174.[181] R. Kubo in Advances in Chemical Physics, Vol. XV: Stochastic

Processes in Chemical Physics (Ed.: K. Schuler), Wiley, NewYork, 1969, p. 101.

[182] Y. Tanimura, J. Phys. Soc. Jpn. 2006, 75, 082001.[183] D. Gamliel, H. Levanon, Stochastic Processes in Magnetic

Resonance, World Scientific, River Edge, 1995.[184] J. H. Freed, G. V. Bruno, C. F. Polnaszek, J. Phys. Chem. 1971,

75, 3385.[185] D. J. Schneider, J. H. Freed in Advances in Chemical Physics,

Vol. LXXIII: Lasers, Molecules, and Methods (Eds.: J. O.Hirschfelder, R. E. Wyatt, R. D. Coalson), Wiley, New York,1989, p. 387.

[186] R. A. MacPhail, R. G. Snyder, H. L. Strauss, J. Chem. Phys.1982, 77, 1118 – 1137.

[187] J. J. Turner, C. M. Gordon, S. M. Howdle, J. Phys. Chem. 1995,99, 17532 – 17538.

[188] F. Sanda, S. Mukamel, J. Chem. Phys. 2006, 125, 014507.[189] S. Woutersen, P. Hamm, J. Chem. Phys. 2001, 114, 6833 – 6840.[190] R. Schweitzer-Stenner, Biophys. J. 2002, 83, 523 – 532.[191] R. Schweitzer-Stenner, F. Eker, Q. Huang, K. Griebenow, J.

Am. Chem. Soc. 2001, 123, 9628 – 9633.[192] S. Woutersen, R. Pfister, P. Hamm, Y. Mu, D. S. Kosov, G.

Stock, J. Chem. Phys. 2002, 117, 6833 – 6840.[193] K. Kwac, M. H. Cho, J. Chem. Phys. 2003, 119, 2256 – 2263.[194] T. L. Jansen, W. Zhuang, S. Mukamel, J. Chem. Phys. 2004, 121,

10577 – 10598.[195] Y. Mu, D. S. Kosov, G. Stock, J. Phys. Chem. B 2003, 107, 5064 –

5073.[196] M. Fayer, personal communication.[197] D. Eisenberg, W. Kauzmann, The Structure and Properties of

Water, Oxford University Press, New York, 1969.[198] The Hydrogen Bond: Recent Developments in Theory and

Experiments, Vol. 1–3 (Eds.: P. Schuster, G. Zundel, C. San-dorfy), North-Holland, Amsterdam, 1976.

[199] P. Wernet, D. Nordlund, U. Bergmann, M. Cavalleri, M.Odelius, H. Ogasawara, L. A. Naslund, T. K. Hirsch, L.Ojamae, P. Glatzel, L. G. M. Pettersson, A. Nilsson, Science2004, 304, 995 – 999.

[200] A. H. Romero, P. L. Silvestrelli, M. Parrinello, J. Chem. Phys.2001, 115, 115 – 123.

[201] E. D. Isaacs, A. Shukla, P. M. Platzman, D. R. Hamann, B.Barbiellini, C. A. Tulk, Phys. Rev. Lett. 1999, 82, 600 – 603.

[202] M. Bernasconi, P. L. Silvestrelli, M. Parrinello, Phys. Rev. Lett.1998, 81, 1235 – 1238.

[203] A. Luzar, D. Chandler, Nature 1996, 379, 55 – 57.[204] S. Woutersen, H. J. Bakker, Nature 1999, 402, 507 – 509.[205] J. H. Guo, Y. Luo, A. Augustsson, J. E. Rubensson, C. Sathe, H.

Agren, H. Siegbahn, J. Nordgren, Phys. Rev. Lett. 2002, 89,137402.

[206] Ultrafast Hydrogen Bonding Dynamics and Proton TransferProcesses in the Condensed Phase (Eds.: T. Elsaesser, H. J.Bakker), Kluwer, Dordrecht, 2002.

[207] R. Laenen, C. Rauscher, A. Laubereau, Phys. Rev. Lett. 1998,80, 2622 – 2625.

[208] J. Stenger, D. Madsen, P. Hamm, E. T. J. Nibbering, T.Elsaesser, Phys. Rev. Lett. 2001, 87, 027401.

[209] R. Rey, K. B. Møller, J. T. Hynes, Chem. Rev. 2004, 104, 1915 –1928.

[210] C. P. Lawrence, J. L. Skinner, J. Chem. Phys. 2002, 117, 5827 –5838.

[211] M. Falk, T. A. Ford, Can. J. Phys. 1966, 44, 1699 – 1714.[212] See Ref. [211][213] S. Yeremenko, M. S. Pshenichnikov, D. A. Wiersma, Chem.

Phys. Lett. 2003, 369, 107 – 113.[214] W. S. Benedicht, N. Gailar, E. K. Plyler, J. Chem. Phys. 1956, 24,

1139 – 1165.[215] S. Woutersen, H. J. Bakker, Phys. Rev. Lett. 1999, 83, 2077 –

2080.[216] H. Graener, G. Seifert, A. Laubereau, Phys. Rev. Lett. 1991, 66,

2092 – 2095.[217] R. Laenen, K. Simeonidis, A. Laubereau, J. Phys. Chem. B

2002, 106, 408 – 417.[218] S. Bratos, J.-Cl. Leicknam, J. Chem. Phys. 1994, 101, 4536 –

4546.[219] A. J. Lock, S. Woutersen, H. J. Bakker, J. Phys. Chem. A 2001,

105, 1238 – 1243.[220] J. Stenger, D. Madsen, P. Hamm, E. T. J. Nibbering, T.

Elsaesser, J. Phys. Chem. A 2002, 106, 2341 – 2350.[221] J. Stenger, D. Madsen, J. Dreyer, P. Hamm, E. T. J. Nibbering, T.

Elsaesser, Chem. Phys. Lett. 2002, 354, 256 – 263.[222] A. Piryatinski, C. P. Lawrence, J. L. Skinner, J. Chem. Phys.

2003, 118, 9672 – 9679.[223] T. Steinel, J. B. Asbury, S. A. Corcelli, C. P. Lawrence, J. L.

Skinner, M. D. Fayer, Chem. Phys. Lett. 2004, 386, 295 – 300.[224] F. Ding, M. T. Zanni, Chem. Phys. 2007, 341, 95 – 105.[225] S. Garrett-Roe, P. Hamm, J. Chem. Phys. 2008, 128, 104507.[226] T. Hayashi, T. L. Jansen, W. Zhuang, S. Mukamel, J. Phys.

Chem. A 2005, 109, 64 – 82.[227] T. L. Jansen, T. Hayashi, W. Zhuang, S. Mukamel, J. Chem.

Phys. 2005, 123, 114505.[228] C. P. Lawrence, J. L. Skinner, J. Chem. Phys. 2003, 118, 264 –

272.[229] F. Csajka, D. Chandler, J. Chem. Phys. 1998, 109, 1125 – 1133.[230] R. Laenen, C. Rauscher, A. Laubereau, J. Phys. Chem. B 1998,

102, 9304 – 9311.[231] I. R. Piletic, D. E. Moilanen, D. B. Spry, N. E. Levinger, M. D.

Fayer, J. Phys. Chem. A 2006, 110, 4985 – 4999.[232] V. V. Volkov, D. J. Palmer, R. Righini, Phys. Rev. Lett. 2007, 99,

078302.[233] D. Cringus, A. Bakulin, J. Lindner, P. Vohringer, M. S.

Pshenichnikov, D. A. Wiersma, J. Phys. Chem. B 2007, 111,14193 – 14207.

[234] Y. R. Shen, The Principles of Nonlinear Optics, Wiley, NewYork, 1984.

[235] Y. R. Shen, Nature 1989, 337, 519 – 525.[236] X. Chen, T. Yang, S. Kataoka, P. S. Cremer, J. Am. Chem. Soc.

2007, 129, 12272 – 12279.[237] M. Sovago, R. K. Campen, G. W. H. Wurpel, M. M�ller, H. J.

Bakker, M. Bonn, Phys. Rev. Lett. 2008, 100, 173901.[238] A. Paarmann, T. Hayashi, S. Mukamel, R. J. D. Mirrer, J. Chem.

Phys. 2008, 128, 191103.[239] J. Park, J. H. Ha, R. M. Hochstrasser, J. Chem. Phys. 2004, 121,

7281 – 7292.[240] J. H. Ha, Y. S. Kim, R. M. Hochstrasser, J. Chem. Phys. 2006,

124, 064508.[241] M. Lima, V. Volkov, P. Foggi, R. Chelli, R. Righini, Biophysics

2005, 45, S85.[242] A. M. Dokter, S. Woutersen, H. J. Bakker, Proc. Natl. Acad.

Sci. USA 2006, 103, 15355 – 15358.[243] H. S. Tan, I. R. Piletic, M. D. Fayer, J. Chem. Phys. 2005, 122,

174501.[244] H. S. Tan, I. R. Piletic, R. E. Riter, N. E. Levinger, M. D. Fayer,

Phys. Rev. Lett. 2005, 94, 057405.





http://dx.doi.org/10.1103/PhysRevA.77.033802





http://dx.doi.org/10.1063/1.1703941

http://dx.doi.org/10.1021/j100691a001

http://dx.doi.org/10.1021/j100691a001

http://dx.doi.org/10.1063/1.443996

http://dx.doi.org/10.1063/1.443996

http://dx.doi.org/10.1021/j100049a011

http://dx.doi.org/10.1021/j100049a011

http://dx.doi.org/10.1063/1.1336807

http://dx.doi.org/10.1021/ja016202s

http://dx.doi.org/10.1021/ja016202s

http://dx.doi.org/10.1063/1.1506151

http://dx.doi.org/10.1063/1.1580808

http://dx.doi.org/10.1063/1.1807824

http://dx.doi.org/10.1063/1.1807824





http://dx.doi.org/10.1063/1.1372183

http://dx.doi.org/10.1063/1.1372183




http://dx.doi.org/10.1038/379055a0

http://dx.doi.org/10.1038/990058






http://dx.doi.org/10.1021/cr020675f

http://dx.doi.org/10.1021/cr020675f

http://dx.doi.org/10.1063/1.1502248

http://dx.doi.org/10.1063/1.1502248

http://dx.doi.org/10.1016/S0009-2614(02)02001-8

http://dx.doi.org/10.1016/S0009-2614(02)02001-8







http://dx.doi.org/10.1063/1.467441

http://dx.doi.org/10.1063/1.467441

http://dx.doi.org/10.1021/jp003158e

http://dx.doi.org/10.1021/jp003158e


http://dx.doi.org/10.1016/S0009-2614(02)00103-3

http://dx.doi.org/10.1063/1.1569474

http://dx.doi.org/10.1063/1.1569474

http://dx.doi.org/10.1016/j.cplett.2004.01.042


http://dx.doi.org/10.1063/1.2883660

http://dx.doi.org/10.1021/jp046685x

http://dx.doi.org/10.1021/jp046685x

http://dx.doi.org/10.1063/1.1525802

http://dx.doi.org/10.1063/1.1525802

http://dx.doi.org/10.1063/1.476656

http://dx.doi.org/10.1021/jp980667w

http://dx.doi.org/10.1021/jp980667w

http://dx.doi.org/10.1021/jp061065c





http://dx.doi.org/10.1038/337519a0

http://dx.doi.org/10.1021/ja073869r

http://dx.doi.org/10.1021/ja073869r


http://dx.doi.org/10.1063/1.2919050

http://dx.doi.org/10.1063/1.2919050

http://dx.doi.org/10.1063/1.1792612

http://dx.doi.org/10.1063/1.1792612

http://dx.doi.org/10.1063/1.2162165

http://dx.doi.org/10.1063/1.2162165



http://dx.doi.org/10.1063/1.1883605

http://dx.doi.org/10.1063/1.1883605



[245] I. R. Piletic, H. S. Tan, M. D. Fayer, J. Phys. Chem. B 2005, 109,21273 – 21284.

[246] A. M. Dokter, S. Woutersen, H. J. Bakker, Phys. Rev. Lett.2005, 94, 178301.

[247] V. V. Volkov, D. J. Palmer, R. Righini, J. Phys. Chem. B 2007,111, 1377 – 1383.

[248] A. Ghosh, M. Smits, J. Bredenbeck, M. Bonn, J. Am. Chem.Soc. 2007, 129, 9608 – 9609.

[249] G. Otting, E. Liepinsh, K. W�thrich, Science 1991, 254, 974 –980.

[250] S. K. Pal, A. H. Zewail, Chem. Rev. 2004, 104, 2099 – 2123.[251] S. Habuchi, H. B. Kim, N. Kitamura, Anal. Chem. 2001, 73,

366 – 372.[252] G. Sposito, N. T. Skipper, R. Sutton, S. H. Park, A. K. Soper,

J. A. Greathouse, Proc. Natl. Acad. Sci. USA 1999, 96, 3358 –3364.

[253] B. Albert, A. Johnson, J. Lewis, M. Raff, K. Roberts, P. Walter,Molecular Biology of the Cell, Garland, New York, 2000.

[254] K. Simons, E. Ikonen, Nature 1997, 387, 569 – 572.[255] R. Mendelsohn, Biochim. Biophys. Acta Biomembr. 1972, 290,

15 – 21.[256] M. R. Bunow, I. W. Levin, Biochim. Biophys. Acta Lipids Lipid

Metab. 1977, 489, 191 – 206.[257] S. F. Bush, H. Levin, I. W. Levin, Chem. Phys. Lipids 1980, 27,

101 – 111.[258] E. Bicknell-Brown, K. G. Brown, W. B. Person, J. Am. Chem.

Soc. 1980, 102, 5486 – 5491.[259] E. Mushayakarara, I. W. Levin, J. Phys. Chem. 1982, 86, 2324 –

2327.[260] W. H�bner, H. H. Mantsch, Biophys. J. 1991, 59, 1261.[261] R. N. Lewis, R. N. McElhaney, W. Pohle, H. H. Mantsch,

Biophys. J. 1994, 67, 2367 – 2375.[262] P. B. Hitchcock, R. Mason, K. M. Thomas, G. G. Shipley, Proc.

Natl. Acad. Sci. USA 1974, 71, 3036 – 3040.[263] R. H. Pearson, I. Pascher, Nature 1979, 281, 499 – 501.[264] H. Hauser, I. Pascher, R. H. Pearson, S. Sundell, Biochim.

Biophys. Acta Rev. Biomembr. 1981, 650, 21 – 51.[265] E. Mushayakarara, N. Albon, I. W. Levin, Biochim. Biophys.

Acta Biomembr. 1982, 686, 153 – 159.[266] A. Blume, W. H�bner, G. Messner, Biochemistry 1988, 27,

8239 – 8249.[267] R. N. Lewis, R. N. McElhaney, Biophys. J. 1992, 61, 63 – 77.[268] V. Volkov, F. Nuti, Y. Takaoka, R. Chelli, A. M. Papini, R.

Righini, J. Am. Chem. Soc. 2006, 128, 9466 – 9470.[269] P. F. Tian, D. Keusters, Y. Suzaki, W. S. Warren, Science 2003,

300, 1553 – 1555.[270] X. M. Yuan, A. K. Downing, V. Knott, P. A. Handford, EMBO

J. 1997, 16, 6659 – 6666.[271] R. W. W. N. Berova, R. W. Woody, K. Nakanishi, Circular

Dichroism. Principles and Applications, 2. Aufl., Wiley, NewYork, 2000.

[272] L. A. Nafle, Annu. Rev. Phys. Chem. 1997, 48, 357 – 386.[273] L. D. Barron, Molecular Light Scattering and Optical Activity,

2. Aufl., Cambridge University Press, Cambridge, 2004.[274] K. K. Lee, C. Joo, S. Yang, H. Han, M. Cho, J. Chem. Phys. 2007,

126, 054505.[275] J. H. Choi, M. Cho, J. Phys. Chem. A 2007, 111, 5176 – 5184.[276] L. D. Barron, L. Hecht, E. W. Blanch, A. F. Bell, Prog. Biophys.

Mol. Biol. 2000, 73, 1 – 49.[277] R. D. Singh, T. A. Keiderling, Biopolymers 1981, 20, 237 – 240.[278] B. B. Lal, L. A. Nafle, Biopolymers 1982, 21, 2161 – 2183.[279] A. C. Sen, T. A. Keiderling, Biopolymers 1984, 23, 1519 – 1532.[280] S. C. Yasui, T. A. Keiderling, J. Am. Chem. Soc. 1986, 108,

5576 – 5581.[281] S. C. Yasui, T. A. Keiderling, Biopolymers 1986, 25, 5 – 15.[282] S. C. Yasui, T. A. Keiderling, G. M. Bonora, C. Toniolo,

Biopolymers 1986, 25, 79 – 89.[283] D. G. Du, Y. J. Zhu, C. Y. Huang, F. Gai, Proc. Natl. Acad. Sci.

USA 2004, 101, 15915 – 15920.[284] K. Schulten, personal communication.[285] R. Tycko, Curr. Opin. Chem. Biol. 2000, 4, 500 – 506.[286] M. A. Findeis, Biochim. Biophys. Acta Mol. Basis Dis. 2000,

1502, 76 – 84.[287] J. Kang, H. G. Lemaire, A. Unterbeck, J. M. Salbaum, C. L.

Masters, K. H. Grzeschik, G. Multhaup, K. Beyreuther, B.M�ller-Hill, Nature 1987, 325, 733 – 736.

[288] D. Burdick, B. Soreghan, M. Kwon, J. Kosmoski, M. Knauer, A.Henschen, J. Yates, C. Cotman, C. Glabe, J. Biol. Chem. 1992,267, 546 – 554.

[289] M. Knauer, B. Soreghan, D. Burdick, J. Kosmoski, C. Glabe,Proc. Natl. Acad. Sci. USA 1992, 89, 7437 – 7441.

[290] T. Luhrs, C. Ritter, M. Adrian, D. Riek-Loher, B. Bohrmann, H.Doeli, D. Schubert, R. Riek, Proc. Natl. Acad. Sci. USA 2005,102, 17342 – 17347.

[291] D. Voronine, D. Abramavicius, S. Mukamel, J. Chem. Phys.2008, 94, 3613 – 3619.

[292] Theory, Modelling and Evaluation of Single Molecule Measure-ments (Eds.: E. Barkai, M. Orrit, F. Brown H. Yang), WorldScientific, Singapore, 2008.

[293] P. Hamm, S. M. Ohline, W. Zinth, J. Chem. Phys. 1997, 106,519 – 529.

[294] P. L. Geissler, C. Dellago, D. Chandler, J. Hutter, M. Parrinello,Science 2001, 291, 2121 – 2124.

[295] H. Maekawa, W. Zhuang, C. Toniolo, S. Mukamel, D. J. Tobias,N.-H. Ge, J. Phys. Chem B (in press).

[296] J. C. Phillips, R. Braun, W. Wang, J. Gumbart, E. Tajkhorshid,E. Villa, C. Chipot, R. D. Skeel, L. Kale, K. Schulten, J.Comput. Chem. 2005, 26, 1781 – 1802.

[297] This is not the case in NMR spectroscopy (see Table 1) in whichthe signal is generated in all directions.


Chemie






http://dx.doi.org/10.1021/jp065886t

http://dx.doi.org/10.1021/jp065886t

http://dx.doi.org/10.1021/ja073130h

http://dx.doi.org/10.1021/ja073130h



http://dx.doi.org/10.1021/cr020689l

http://dx.doi.org/10.1021/ac0007276

http://dx.doi.org/10.1021/ac0007276



http://dx.doi.org/10.1038/42408

http://dx.doi.org/10.1016/0005-2736(72)90047-8

http://dx.doi.org/10.1016/0005-2736(72)90047-8

http://dx.doi.org/10.1016/0005-2760(77)90138-2

http://dx.doi.org/10.1016/0005-2760(77)90138-2

http://dx.doi.org/10.1021/ja00537a011


http://dx.doi.org/10.1021/j100210a016

http://dx.doi.org/10.1021/j100210a016

http://dx.doi.org/10.1016/S0006-3495(94)80723-4



http://dx.doi.org/10.1038/281499a0

http://dx.doi.org/10.1016/0005-2736(82)90107-9

http://dx.doi.org/10.1016/0005-2736(82)90107-9



http://dx.doi.org/10.1016/S0006-3495(92)81816-7




http://dx.doi.org/10.1093/emboj/16.22.6659

http://dx.doi.org/10.1093/emboj/16.22.6659


http://dx.doi.org/10.1016/S0079-6107(99)00017-6

http://dx.doi.org/10.1016/S0079-6107(99)00017-6

http://dx.doi.org/10.1002/bip.1981.360200117

http://dx.doi.org/10.1002/bip.360211106








http://dx.doi.org/10.1016/S1367-5931(00)00123-X

http://dx.doi.org/10.1038/325733a0




http://dx.doi.org/10.1063/1.473392

http://dx.doi.org/10.1063/1.473392





Wei Zhuang, Tomoyuki Hayashi, and Shaul Mukamel*

Documents